AIニュース 2026-07-29

Concept-based Visual Counterfactual Explanations with Diffusion Models

Visual counterfactual explanations aim to answer "what minimal change to this image would flip the model's prediction?", and are increasingly important as vision models are deployed in safety-critical domains (e.g., medicine). Existing diffusion-based methods can produce realistic edits, but they rely on external classifiers that must work reliably on noisy images, which makes them fragile and hard to deploy for robust explanations. We introduce C-VCE, a new diffusion framework that builds the classifier directly into the generative model via a concept bottleneck layer, so that counterfactuals are guided by human-interpretable features (concepts) instead of a separate noise robust classifier that works with pixel-level edits. Our model lets users to toggle on/off semantic concepts during sampling, then minimally adjusts relevant image regions, while preserving the rest of the image, respecting feature correlations. To keep edits small and controlled, we add a simple probabilistic regularizer that balances "change the prediction" against "stay close to the original", plus a gradient-based mask that confines modifications to the most relevant regions. On benchmarks such as CelebA, C-VCE matches or improves flip rates while producing counterfactuals that are visually closer to the input and less distorted than baselines that depend on separate noisy-image classifiers. These properties make C-VCE a practical tool for vision systems where users need concrete "what-if" images without having to trust an additional, noise-robust classifier. More broadly, our results suggest that exposing and controlling an internal concept layer is a promising way to make powerful generative models easier to understand and safer to use.

SeT-Diff: Towards Semantic Foundation Models for HPC Telemetry and Time-Series

Data centers and their compute nodes require accurate and flexible digital twins capable of modeling the complex interplay of workloads, environmental parameters, and physical metrics. Current machine learning approaches for HPC and its telemetry typically rely on a static subset of anonymous, fixed-position sensor variables tailored to single tasks. Consequently, these models become obsolete when target tasks change or sensor metrics vary. We propose SeT-Diff, the first foundational model for compute node telemetry and time-series. Unlike rigid architectures, our diffusion-based approach conditions the generative process on each sensor's semantic description, decoupling the system dynamics from the structure of the dataset. Experiments on a real-world supercomputer dataset demonstrate a Mean Absolute Error (MAE) of 0.0470 on reconstruction tasks. SeT-Diff exhibits zero-shot permutation stability, maintaining accuracy with negligible degradation even when sensors are shuffled. A single pre-trained model effectively performs data imputation, forecasting, and virtual sensing - achieving a 0.033 MAE in thermal inference - making SeT-Diff an effective data-driven digital twin for HPC systems.

QFoldAgent: An Autonomous Quantum Optimization Multi-Agent System for Protein Structure Prediction

Hybrid quantum-classical protein structure prediction depends strongly on Hamiltonian penalty weights, yet existing lattice-based workflows typically fix these coefficients by hand and evaluate only very short fragments in simulation. We present QFoldAgent, a closed-loop multi-agent framework for 5-residue tetrahedral-lattice folding in which a design agent proposes sequence-conditioned penalties, a VQE-based quantum-classical pipeline optimizes the resulting Hamiltonian under Qiskit Aer noise, and a feedback agent uses energy-landscape diagnostics and MolProbity validation signals to refine penalties across cycles. Ground-truth metrics such as RMSD are never exposed to the agents and are used only for evaluation. We study the framework on two complementary datasets: 55 QDockBank-derived fragments with known structures and 100 coverage-optimized unseen sequences. On the QDockBank benchmark, QFoldAgent reduces median RMSD from 3.64 \AA{} to 3.20 \AA{}, with the largest gains on the hardest targets. On unseen sequences, the closed loop raises structural validity from 87.5% to 98.7%, recovers 87% of initially invalid cases, and the strongest controller improves cycle-3 energy on 87% of sequences while maintaining 96% Ramachandran-favored geometry. These results show that iterative agent control can systematically improve optimization behavior and reduce failure cases in a 5-residue quantum setting.

Same Question, Different Answers: Evaluating LLM Reliability Beyond Accuracy

Large language models (LLMs) often achieve strong accuracy on benchmarks, yet it remains unclear how reliably they apply this knowledge when the same question is phrased in different but equivalent ways. In this work, we study how model answers change under meaning-preserving paraphrases across factual question answering and mathematical reasoning tasks. Across four benchmarks and 13 models, we find that model outputs frequently depend on the exact wording of the prompt. While overall accuracy typically changes only modestly across paraphrases, instance-level behavior is far less stable: for many questions, models alternate between correct and incorrect answers depending on phrasing, with mismatch rates reaching more than 23%. Conditioning on questions that are answered correctly in their original form reveals even larger failures measured by answer flip rates, showing that single-prompt correctness is often a poor indicator of reliability. At the same time, we find that models often produce a correct answer for at least one paraphrase of a question, suggesting that the underlying knowledge is present but inconsistently retrieved. Building on this observation, we show that a simple self-paraphrasing strategy can partially recover this latent knowledge and improve performance at inference time. Together, these findings suggest that standard accuracy metrics can mask substantial instability, and that evaluating consistency across equivalent inputs provides a clearer picture of LLM reliability.

DeepLens Diagnosis Agent: Agentic Workflow Design Lets a Small Reasoning Model Compete with Frontier LLMs

Medical diagnosis is a multi-stage process: extract facts, consult knowledge, generate a differential analysis, and select the best diagnosis with explanations. Frontier LLMs are strong generalists, but single-shot prompting often yields brittle diagnostic reasoning. We present the DeepLens Diagnosis Agent, a five-stage harnessing pipeline (combining model capabilities with disciplined process constraints) centered on a small medical reasoning model (JSL Medical Small 7B v2) and retrieval-augmented generation (RAG). The pipeline enforces structured clinical extraction, disciplined retrieval, constrained candidate generation, explicit evidence triangulation, and an auditable final decision. On the 915-case DiagnosisArena benchmark, the agent achieved 60.14% top-1 diagnostic accuracy, the highest among small and medium-sized models. The same model without the agent workflow achieved 23.99%, a +36-point gain from workflow design alone, despite 88.2% on standard medical benchmarks, showing that diagnostic reasoning under uncertainty requires more than knowledge recall. The agent costs USD 0.0072 per case (24K tokens on A100) with 24-second latency, 35-45% cheaper than Claude Sonnet 4.5 (USD 0.0110) and Gemini 3.1 Pro (USD 0.0128) while outperforming them by +9.70pp and +9.17pp. Harnessing can also correct frontier model failures; workflow constraints can outweigh parameter count or API cost. Beyond aggregate accuracy, the pipeline produces structured intermediate artifacts that make each stage inspectable and support error localization. These properties support high-stakes settings where traceability, reproducibility, and auditable evidence matter alongside benchmark performance.

MIITA: Memory-Induced Inference-Time Adaptation for Continual Learning with Small Language Models

Continual learning (CL) is essential for small language models (SLMs) to adapt to evolving real-world needs in resource-constrained deployments. However, directly updating their limited parameter space causes catastrophic forgetting. While memory-based methods naturally address this by decoupling knowledge retention from parameters, existing approaches designed for large language models (LLMs) rely on abundant storage and strong in-context reasoning that SLMs lack. To address these challenges, we propose MIITA, a Memory-Induced Inference-Time Adaptation framework for supervised CL under constrained storage. MIITA stores supervised experiences as compact correction-direction prototypes with semantic anchors, and retrieves them at inference time using semantic and uncertainty-based cues. The retrieved directions are applied through gated temporary hidden-state adaptation, enabling non-destructive reuse of past supervision without backbone updates, prompt extensions, or test-time backpropagation. A local theoretical analysis links this design to first-order loss reduction, uncertainty-guided retrieval, and directional coverage for retaining old-stage knowledge. Extensive experiments across diverse supervised CL settings show that MIITA consistently improves final performance and mitigates forgetting under fixed memory budgets.

Codifying the Judge: Scalable Evaluation via Program Distillation

LLM-as-a-judge has become the standard for automated evaluation, but it suffers from high cost, significant latency, and opaque decisions -- limitations that undermine its scalability and reliability. We address these with a simple, efficient alternative: program distillation. Instead of prompting an LLM at the evaluation time, we distill its decision logic into a committee of programs that score candidates directly. These programmatic judges offer transparency, are easily inspected or edited, and eliminate per-sample API costs. Building on this notion, we introduce PAJAMA, a system that synthesizes programs as judges, aggregates their decisions into a joint verdict, and incorporates a fallback mechanism to selectively escalate low-confidence cases to an LLM. Across five datasets and four model families, we show that programmatic judges can match the performance of a 13B-size LLM judge. When using program outputs as routing signals, PAJAMA improves both accuracy and throughput and advances the Pareto frontier. Beyond evaluation, programmatic judges produce cheap and effective reward signals: on RewardBench, a reward model distilled from programs' verdicts outperforms one trained on a proprietary LLM's labels at two orders of magnitude lower API cost.

SF-AMS: Strategic Forgetting for Structured Memory in LLM Agent

Managing long-context dependencies remains a primary bottleneck in LLM agents, as redundant and irrelevant information can degrade multi-step reasoning. Strategic Forgetting for Agent Memory Systems (SF-AMS) is proposed as a framework for maintaining compact high-utility memory by modeling the long-term importance of memory units. SF-AMS replaces static retrieval and heuristic decay with a utility-driven survival mechanism that updates memory importance from usage redundancy and temporal signals, inducing a hierarchical memory structure that prioritizes stable entity-consistent information while filtering noise. On top of this, Composite Importance Scoring integrates semantic and entity level signals to improve retrieval robustness. Experiments on LoCoMo and LongMemEval-s show consistent gains over strong state of the art baselines including LightMem MemO and A-Mem. The largest improvement appears in multi-hop reasoning under Qwen2.5-7B where SF-AMS achieves plus 9.65 F1 over the strongest baseline followed by temporal reasoning under GPT-4o-mini plus 6.91 F1 and open-domain tasks plus 6.53 F1 demonstrating strong cross backbone generalization. These results show that modeling memory importance as a dynamic utility signal is critical for reliable long-context reasoning.

Synthetic Scenario Generation for Evaluation of Industry 4.0 Agents

Industrial agent benchmarks require realistic evaluation scenarios that integrate telemetry, failure modes, maintenance records, and domain standards. However, existing benchmarks such as AssetOpsBench rely on manually authored scenarios and cover a limited set of asset classes. We extend AssetOpsBench with a Smart Grid Transformer asset class and four IEC-grounded diagnostic tools for health-index prediction, dissolved-gas analysis, winding-temperature assessment, and load-profile assessment. We further introduce ScenarioGeneratorAgent, a pipeline for synthetic industrial-agent scenario generation. The pipeline constructs evidence-grounded asset profiles, allocates coverage-aware scenario budgets across operational domains, and generates candidates through a hybrid validation-and-repair loop that enforces schema validity, tool reachability, physical plausibility, standards alignment, and deduplication. To improve scalability, we apply two-level caching, parallel focus-group generation, thread-pool offloading, batched LLM calls, and early rejection filtering. On Smart Grid Transformer scenario generation, these optimizations reduce end-to-end runtime by $8\times$ for 50 scenarios while preserving quality, achieving a composite quality score of $74.2 \pm 1.9$ compared with $73.8 \pm 3.0$ for the unoptimized baseline. These results show that standards-grounded synthetic scenario generation can efficiently expand industrial-agent benchmarks without sacrificing scenario quality.

Loss-Aware Feature-Map Pruning in Convolutional Neural Networks Using Multi-Armed Bandits

Convolutional neural networks often contain redundant feature maps that increase storage and inference cost. This paper presents a loss-aware feature-map pruning framework using multi-armed bandits. Feature-map pruning is structured because it removes complete convolutional output channels and their producing filters rather than isolated scalar weights. Each candidate feature map is treated as an arm. At each play time, one map is temporarily masked and evaluated on a sampled mini-batch; the map is then restored and the observed loss change is converted into a safe-removal reward. After a fixed play budget, candidate maps are ranked by learned scores and the top-k maps are permanently removed with their filters, biases and corresponding next-layer input-channel kernels. The study evaluates UCB1 and Thompson Sampling, compares them with direct/oracle-style evaluation on LeNet/MNIST, and extends the evaluation to MNIST, CIFAR-10, CIFAR-100, SVHN, CUB-200-2011 and Oxford Flowers 102. Results show that UCB1 and Thompson Sampling preserve accuracy close to unpruned models while removing feature maps and reducing convolutional computation. Friedman and Nemenyi tests show that UCB1 obtains the highest mean rank, followed by Thompson Sampling; both significantly outperform greedy and magnitude-based pruning while remaining statistically comparable to the original unpruned model.

DSTFView: Multi-View Cloud-Edge Workload Forecasting with Dual-Input Spatio-Temporal-Frequency Modeling

With the widespread deployment of edge-side AI inference, edge platforms are increasingly required to support latency-sensitive, highly concurrent, and reliability-critical applications. However, existing methods often struggle to balance multidimensional feature modeling and forecasting efficiency in collaborative cloud-edge environments. To address this issue, we propose DSTFView, a dual-input spatio-temporal-frequency multi-view workload forecasting framework for collaborative cloud-edge environments. It jointly models closeness and period dependencies and extracts spatial, temporal, and frequency-domain dependencies. Besides, it designs an adaptive fusion mechanism and adjusts the contribution of each view to capture abrupt changes. Experimental results on the CPU and TP datasets demonstrate that DSTFView consistently outperforms representative baselines across multiple forecasting horizons and evaluation metrics.

MedLoCoMo: A Long-Context Multi-Session Medical Dialogue Benchmark for Large Language Models

MedLoCoMo is a Medical Long-Context Memory benchmark for patient-specific clinical reasoning over multi-admission medical dialogue. Existing medical QA benchmarks largely test short context knowledge or single document grounding, leaving open whether LLMs can use, connect, and abstain over longitudinal patient histories. We build MedLoCoMo from deidentified MIMIC-IV and MIMIC-IV-Note records by constructing admission-level clinical packets, synthesizing grounded doctor-patient conversations, and generating evidence linked QA items over single-admission, cross-admission, and adversarial unanswerable settings. The benchmark contains 100 patient timelines averaging 1,669.8 turns, 29.7 sessions, and 74,512.2 tokens per conversation. Across the evaluated baselines, cross-admission reasoning is consistently harder than localized evidence use, even when models have long context windows or use external memory or retrieval methods. The code and MedLoCoMo benchmark release is available at https://github.com/leozzy13/MedLoCoMo for use and reproducibility.

Keyword Matters: Unveiling the Energy Sensitivity of On-Device LLM Prompting

Large Language Models (LLMs) are increasingly deployed on mobile and embedded devices to improve privacy and reduce network latency. Yet on-device inference faces a fundamental constraint: high energy consumption on battery-powered, resource-limited hardware. While model compression and runtime acceleration have been widely studied, the effect of \emph{prompt design} on energy efficiency remains underexplored. This paper presents an empirical study of the relationship between prompt wording and energy consumption for on-device LLMs. Using real power measurements collected on a smartphone, we quantify how linguistic features, particularly imperative keywords and instruction structure, affect decoding length and total energy. Our results show consistent energy differences across verbs and tasks, indicating that prompt engineering is a lightweight lever for improving energy efficiency.

Execution-Grounded Security Testing for Coding Agents in Software Engineering Pipelines

Coding agents are increasingly integrated into system operations, where their tool use can directly modify project artifacts, execution environments, and the underlying system. For example, if a coding agent inserts a hook into a system startup or configuration script, that change can persist after the interaction, be triggered later, and abuse delegated user or system privileges to modify the system. This makes security testing a system problem: the key question is not only what the agent says, but what it actually does to the surrounding environment. We present an execution-grounded red-team testing framework for probing this execution-layer security boundary using observable sandbox evidence, including tool invocations, runtime traces, and file-system diffs. Our framework embeds target unsafe operations into routine software engineering workloads, including unit testing, regression testing, crash reproduction, and validation, and uses an execution oracle to guide refinement when an initial probe is rejected or fails. Across multiple agent frameworks and model backbones, our red-team workload reformulation substantially increases verified unsafe execution, reaching 73.61% on code carriers and 53.93% on text carriers. These results show that coding agents in system operations remain insecure under task disguise: once risky intent is hidden inside plausible engineering tasks, the agent can be induced to carry out unsafe actions on the surrounding system. More broadly, coding agents in system operations still demand stronger security testing and safeguards.

Reference Feature Atlases for Mechanistic Auditing of Language Models

Auditing a new language model usually means relearning and reinterpreting its internal features from scratch. We propose a reference feature atlas: a sparse feature library trained once on a reference panel and reused for new targets, which attach by fitting only a linear decoder. This yields two complementary views. The atlas channel reads the target on already interpreted panel features, providing a stable coordinate system across models. The residual channel learns features only from what the atlas fails to reconstruct, making "outside the reference panel" an explicit audit signal. We train leave-one-out atlases over five 7-9B instruction-tuned models and audit held-out Mistral and Qwen targets. On three controlled LoRA hidden objectives injected into both targets, the residual channel makes the planted mechanism perfectly controllable at runtime while matched controls stay unaffected and recovers the planted objective as the top-ranked latent across both targets; on Mistral, where the per-target SAE and pairwise crosscoder baselines are retrained for a head-to-head benchmark, both baselines fail to do so. On Qwen-2.5, the same channel additionally reveals a panel-relative political-framing cluster; steering it shifts the audited framing metrics while out-of-domain controls remain unchanged.

SCAIR: Schema-Conditioned Agentic Iterative Reasoning for Enterprise Knowledge Graphs

Knowledge Graph-based Retrieval-Augmented Generation (KG-RAG) enables natural language interaction with structured enterprise knowledge, yet existing agentic approaches that perform well on public benchmarks often fail to generalize to real-world enterprise Knowledge Graphs (KGs), which are dense, schema-driven, and operationally constrained. To address these limitations, we propose SCAIR (Schema-Conditioned Agentic Iterative Reasoning), a training-free framework that integrates structured planning with controlled iterative reasoning by injecting schema-conditioned structural priors and enforcing schema-aware traversal during multi-hop reasoning. Experiments on an enterprise-oriented benchmark constructed from a real-world Configuration Management DataBase (CMDB) demonstrate that SCAIR substantially improves performance over existing KG-RAG methods. Crucially, our study highlights that reliable enterprise graph reasoning cannot rely on generic agentic designs; instead, it must explicitly incorporate the target domain's structural and operational constraints into the reasoning process. We demonstrate that by aligning agent design with business logic, substantial performance gains can be achieved without the need for costly model retraining.

Schema-Aware Localisation (SAL): Live Schema Grounding and Hallucination Validation for Oracle NL2SQL

Large language models can generate fluent SQL from natural language, but on real enterprise Oracle databases they frequently fail at execution time: columns and aliases are hallucinated and dialect-specific syntax is missed, leading to ORA-00904 invalid-identifier errors. In this setting, failures are primarily due to missing schema grounding: the model cannot know which tables and columns actually exist. This paper introduces Schema-Aware Localisation (SAL), a lightweight middleware layer for Oracle NL2SQL that requires no model retraining. SAL queries Oracle's USER_TAB_COLUMNS catalog to build a live schema map, selects a relevant table subset for each question (falling back to the full schema for multi-table queries), and injects this ground-truth context into the LLM prompt. Generated SQL is then checked by the Hallucination Index (Hidx), which validates every alias.column reference against the live catalog, automatically rewrites predictable prefix errors, and otherwise triggers a structured retry with itemised corrections. We evaluate SAL on 500 TPC-H natural language questions executed against a live Oracle Autonomous Database 23c instance using GPT-4o-mini. Without any schema grounding, execution-grounded truth (EGT; executes and matches the reference result set) is 2.2% (12/500). A hand-written static schema hint brings EGT to 62.0%. SAL, with no manual schema curation, achieves 62.6% EGT (96% simple, 95% medium, 40.7% complex) while reducing execution failures from 97.6% to 2.6%.

PhononBench-MP40: a spectrum-resolved benchmark dataset for phonon stability

Imaginary phonon modes remain a practical bottleneck in computational materials screening because otherwise plausible structures can be locally dynamically unstable under a chosen workflow. Here we present PhononBench-MP40, a spectrum-resolved benchmark dataset of Materials Project-derived crystals for workflow-defined phonon stability. The dataset starts from 47,969 MP40 workflow tasks and provides 46,899 completed records with paired stability labels and local phonopy YAML spectra, including 16,683 Stable records and 30,216 completed-phonon unstable records. A further 1,067 relaxation failures are reported separately rather than merged into the completed phonon denominator. The release centers on the local YAML spectrum: the stability label, the lowest sampled frequency and any threshold-dependent relabeling are derived from that spectrum. The dataset is openly available through Science Data Bank at https://doi.org/10.57760/sciencedb.38735. A companion GitHub repository provides the calculation code and lightweight access utilities. PhononBench-MP40 provides an auditable reference for workflow-defined stability classification, minimum-frequency analysis, threshold studies and failure-aware triage, while keeping the reference workflow, data schema and interpretation boundaries explicit.

Too much evidence, too little time: From text to actionable recommendations through multi-objective evidence reasoning

Evidence-based clinical decision making requires specialists to identify, evaluate and synthesize relevant scientific literature. However, PubMed searches for complex clinical cases often return hundreds of publications that cannot be reviewed manually under time constraints. This study proposes SCEPTER (Single-Case Evidence-driven PubMed-To-rEcommendation Reasoner), a framework for transforming clinical case descriptions into evidence-based recommendations. SCEPTER combines PubMed retrieval, PubMedBERT semantic ranking, large language model (LLM)-based claim extraction, evidence-level weighting, contradiction detection, consensus analysis and multi-objective Pareto claim selection. The framework generates structured evidence syntheses and grounded actionable recommendations. A Paper Q&A module further enables interactive exploration of selected publications. The proposed framework introduces multi-objective reasoning model that integrates literature support, contradiction analysis and interactive literature interrogation into a unified clinical decision-support pipeline. Evaluation on 150 case studies demonstrated that the framework reduced an average search space of 576 papers to 53 retained papers, 7 Pareto-optimal claims and 3 final recommendations, corresponding to an overall compression ratio of 192:1. Despite this reduction, the retained evidence maintained high diversity (entropy=0.901). The ablation study showed that Pareto-based selection increased evidence diversity and recommendation utility compared with conventional ranking approaches.

Temporal Context Reinstatement Drives Episodic-Like Order Memory in Long-Context Language Models

Human episodic memory supports the retrieval of experiences that unfold over extended timescales, yet the computational mechanisms underlying this ability remain debated due to the limited mechanistic accessibility in long-term memory experiments in humans. Long-context LLMs may offer promising ways to reveal plausible computational mechanisms that drive this type of retrieval. Here, we investigate whether and how LLMs capture the core behavioral signatures of episodic memory via a temporal order memory task. Using a new dataset of human behavior based on memory of a full-length novel, we show that models exhibit the same characteristic distance effect observed in humans on this task. We next apply long-context mechanistic interpretability analyses to uncover how models solve this task, and find that model performance relies on a one-dimensional temporal code that is reinstated during retrieval by a single time-reinstatement attention head. These findings support temporal context reinstatement as an important mechanism for episodic-like temporal-order memory in LLMs, offering new insights into how temporal aspects of long-term episodic memory may be instantiated in both artificial and biological systems.

cMoLLM at Scale: Horizontal Scaling Laws for Mixture-of-LLMs

Scaling large language models (LLMs) has driven their success, yet dense Transformers couple capacity and computation: every parameter is activated for every token, making training and inference costs grow linearly with model size-a critical bottleneck as models approach trillion-parameter regimes. We aim to scale capacity through MoE-style mixture throughout the LLM pipeline rather than only the FFN. Prior pipeline-level approaches include ParaScale, which introduces virtual tokens and parallel streams but incurs substantial overhead and suffers from homogenized routing and gradient collapse, and AltUp, which uses an auxiliary prediction branch but offers limited adaptivity and slow convergence. We establish that MoE-style mixture layers can be reformulated as variable-kernel dynamic convolutions, where each expert corresponds to a $1{\times}1$ convolutional kernel and routing implements input-conditioned kernel aggregation. Building on this equivalence, we introduce cMoLLM: a convolutionally gated mixture-of-LLMs that routes over end-to-end streams through fully differentiable dynamic convolution. In GPT-2-style models trained on FineWeb, cMoLLM improves language modeling perplexity and downstream GLUE and SQuAD accuracy under matched compute, with better stream utilization, more stable optimization, and favorable scaling compared to ParaScale- and AltUp-style baselines.

HeraSys: Collaborative Serving of Multiple LLM Workflows via Fine-Grained End-to-End Optimization

The proliferation of Large Language Models (LLMs) has shifted serving systems from processing isolated requests to orchestrating high-concurrency, multi-tenant agentic workflows. However, existing solutions typically prioritize intra-workflow optimization, largely neglecting the significant potential for inter-workflow optimization. In this paper, we propose HeraSys, an LLM serving system designed to optimize the end-to-end performance of concurrent workflows. Through fine-grained orchestration, HeraSys eliminates cross-workflow computational redundancy via structural node merging and reuse. Furthermore, HeraSys introduces a load-aware joint scheduling policy that dynamically manages execution order by evaluating both inter- and intra-query priorities. By integrating a resource skewing mechanism with adaptive batching and pipeline decomposition, HeraSys effectively mitigates tail latency while maintaining low average latency, thereby substantially improving system throughput. Extensive experiments demonstrate that HeraSys reduces P99 latency by up to 2.17$\times$ and increases serving throughput by up to 1.85$\times$ under strict latency guarantees.

Multi-Objective Structured Pruning of LLMs for Latency and Model Size Optimization

Large Language Models (LLMs) have achieved widespread adoption because of their strong reasoning and query-response capabilities. However, deploying them in embedded and edge computing environments remains challenging because of strict latency, memory, and energy constraints. Their large parameter counts and computational demands hinder efficient execution on resource-constrained platforms. Although model pruning has emerged as a viable solution for reducing scale while preserving performance, jointly optimizing layers, attention heads, and Multi-Layer Perceptron (MLP) dimensions remains highly complex. Exhaustively exploring this combined design space is computationally expensive and often leads to local optima or unstable configurations. To address these limitations, we propose a hardware-aware, multi-objective structured pruning framework. The proposed two-stage method explicitly targets latency and model size for efficient deployment on edge devices. In the coarse-grained stage, multi-objective depth pruning removes entire attention and MLP blocks to reduce computational load and memory usage. In the subsequent fine-grained stage, Parallel Bayesian Optimization (PBO) searches for the optimal layer-wise pruning ratios for pruning under latency constraints, while importance-based strategies rank the specific components to be pruned within each layer's allocated budget. Experimental results show that our approach reduces model complexity with minimal impact on commonsense reasoning tasks and zero-shot performance. Our method achieves a favorable trade-off among accuracy, latency, and model size, making it suitable for edge deployment. Across multiple LLMs at 37.5% and 50% pruning ratios, the proposed approach achieves better performance on commonsense reasoning tasks than existing methods while significantly reducing inference cost.

Source-Aware Reranking for Retrieval-Augmented Generation: A Reliability Prior Approach

Standard Retrieval-Augmented Generation pipelines rank retrieved documents by semantic similarity alone, without accounting for source provenance or credibility. This work evaluates a simple and interpretable modification to RAG retrieval ranking that incorporates domain-informed source reliability priors. Each document is assigned a prior lambda(s) based on its source type, and retrieval scores are reweighted using score(q, d) = sim(q, d) * lambda(s). The framework is evaluated against a similarity-only baseline on a 120-document health-domain corpus. In this controlled setting, source-aware reranking improves Precision@5 from 0.48 to 0.72 and reduces average adversarial document retrieval under the evaluated threat model, where low-credibility sources are identifiable via metadata. All experiments were executed on Rosie, the high-performance computing cluster at the Milwaukee School of Engineering, which provided the GPU-accelerated infrastructure necessary to run the full experimental pipeline reliably and reproducibly. These results suggest a potential mitigation strategy for source quality degradation in RAG pipelines, within the limits of the experimental setup described.

The Scaffold Effect in Coding Agents: Harness Choice as a Hidden Variable in Coding-Agent Evaluation

Public leaderboards for coding agents typically rank systems by model name and pass rate, while the surrounding harness (the scaffold that issues tools, manages context, and decides when to stop) is often under-specified. Model-to-model comparison is valid when the harness is fixed; when it varies, performance and efficiency conflate model and scaffold effects. We evaluate Qwen 3.6 Plus and MiniMax M2.5 across three open-source harnesses (Goose, OpenCode, OpenHands-SDK) on a stratified 50-task subset of Terminal-Bench Pro. Harness choice induces up to a 40x difference in tokens per solved task, while paired within-model pass-rate differences remain 0-8 percentage points (95% paired-task bootstrap CIs include zero except for the largest gap). Failure fingerprints replicate across models (REASON for Goose, VERIFY/MAX_TURNS for OpenHands-SDK, idle-loop/TIME for OpenCode), indicating harness-level biases that are largely model-independent. For human-centered coding-agent evaluation, model name alone is an incomplete comparison unit: harness-model pairs determine real-world cost, latency, and oversight burden; no-action turns are a per-task wait tax, not just a token tax. We therefore recommend selecting harness-model pairs by pass rate under token/latency budgets, and reporting token usage, latency, and full harness specifications alongside any model comparison. We release anonymized configs, raw trial logs, aggregated snapshots, and analysis scripts.

MM-ShiftKV: Decode-Aware Prefill-Stage KV Selection for Multimodal Large Language Models

Key-Value (KV) caching is essential for efficient inference in multimodal large language models (MLLMs), yet its memory footprint grows linearly with context length and becomes a major bottleneck due to the large number of visual tokens. Recent prefill-stage KV selection methods estimate KV importance from prefilling statistics, implicitly assuming that prefilling-time queries are representative of those encountered during decoding. We show that this assumption breaks down in multimodal inference, where decoding-time queries exhibit substantially larger variance than prefilling-stage representations, leading to unstable KV importance estimation under tight cache budgets. As a result, small ranking errors can disproportionately discard semantically critical visual tokens and degrade grounding and reasoning performance. We propose MM-ShiftKV, a training-free, decode-aware and strictly prefill-only KV selection method. MM-ShiftKV approximates decoding-time query behavior during prefilling by constructing variance-expanded query proxies and estimates prompt KV importance based on their aggregated attention mass. Experiments on multimodal benchmarks demonstrate that MM-ShiftKV consistently outperforms existing methods under strict KV-cache budgets. Our code is available at https://github.com/zjuDBxAI/MM-ShiftKV.

TriSP: Tri-Signal Structured Pruning for Large Language Models

Large language models (LLMs) achieve strong performance across diverse tasks but their deployment is constrained by the memory and compute cost of their parameters. Structured pruning addresses this by removing entire structures such as attention heads and Multi-Layer Perceptron (MLP) neurons to produce smaller dense models that run efficiently on standard hardware. However, existing methods rely on either gradient-based importance estimation, which is memory-prohibitive, or activation-based statistical proxies, which do not directly measure the effect of removal on the loss. Furthermore, the interaction between the importance criterion and the post-pruning recovery strategy has not been systematically studied. We propose TriSP (Tri-Signal Structured Pruning), an importance metric that combines weight magnitude scaled by activation norm with first-order gradient sensitivity via a geometric mean, producing a channel-level score that captures both structural and loss-sensitivity signals. Combined with adaptive per-layer budget allocation and low-rank adaptation (LoRA) recovery, TriSP achieves the lowest perplexity and highest zero-shot accuracy across all tested configurations, reaching 6.80 WikiText-2 perplexity at 20% pruning on LLaMA-7B. Inference throughput improves by 82% at 50% pruning, while still maintaining competitive performance.

ParBench: A Benchmark for Reliable Evaluation of LLM Parallel Code Translation

Modern compute-intensive software must migrate across a changing ecosystem of accelerators, programming APIs, compiler stacks, and portability layers, including CUDA, OpenMP, OpenCL, and OpenMP target offload. Large language models and autonomous coding agents are increasingly proposed for such migration, but the field lacks reliable ways to measure whether they preserve the low-level parallel semantics that make translations behaviorally valid, including thread indexing, synchronization, memory management, host-device coordination, and API-specific execution structure. We present ParBench, a kernel-centric benchmark framework for evaluating LLM-based parallel API translation under executable, reproducible conditions. ParBench fixes the surrounding build, run, and verification infrastructure through declarative benchmark specifications and asks models to translate only the computational kernels. It draws on multiple open-source HPC suites and covers representative cross-API translation directions among CUDA, OpenMP, OpenCL, and OpenMP target offload. To test whether success reflects robust translation rather than surface-form memorization, ParBench includes AST-driven, intended behavior-preserving, baseline-validated source augmentation. Evaluations on state-of-the-art open and proprietary LLMs show persistent barriers to reliable parallel code translation, including direction asymmetry, multi-file coordination, incomplete API adaptation, and uneven robustness to source-level perturbations. Code is available at https://github.com/Scientific-Computing-Lab/ParBench.

Lexical discovery in unknown environments orchestrated by Large Language Models

Populations of autonomous agents deployed in unknown environments (e.g. planetary or deep-sea exploration) must develop shared vocabularies to refer to entities that have no name in any human language. We propose the Neuro-Symbolic Lexical Discovery (NSLD) framework, in which a population of LLM-based agents plays a referential game over out-of-distribution visual referents, autonomously self-organising a shared alien lexicon. Each agent combines a frozen CLIP vision encoder with a private FAISS vector index and a text-only LLM. Crucially, discovered alien words are anchored to natural language via semantic proximity in the embedding space, enlarging the human vocabulary with new perceptually grounded words. Consensus is reached in simulations with populations of up to twenty agents and ten visual referents. Convergence dynamics are characterised through three analytical models achieving R^2 > 0.95, representing a first step towards pre-deployment planning in autonomous exploration missions.

Structure Over Scale: Schema-Constrained Causal Graphs for RAG

Graph-based retrieval-augmented generation (GraphRAG) grounds answers in structured knowledge, but current systems extract entities and relationships exhaustively, producing graphs whose size and construction cost scale with corpus length rather than with the reasoning a query requires. We introduce HCG-RAG (Hierarchical Causal Graph RAG), which replaces open-ended extraction with schema-constrained causal graphs: an automated pipeline distills a corpus into a fixed, typed vocabulary of causal variables and materializes a compact two-tier graph over it. Our schema-constrained graphs match entity-relation baselines on answer quality at a fraction of the cost: 3-20x fewer nodes, 8x-135x fewer build-time LLM calls than the most LLM-intensive baseline (MS-GraphRAG), and graphs compact enough for a domain expert to audit, correct, and extend. On medical and clinical benchmarks, including a neurologist-validated epilepsy dataset, HCG-RAG matches or exceeds the best entity-relation systems. An ablation isolates the causal graph as a structured retrieval filter, contributing +6 percentage points (pp) over embedding-only retrieval. Across all domains with discoverable hierarchical causal structure, only methods imposing higher-level organization outperform flat entity-relation retrieval, indicating that what is placed in the graph matters more than how many nodes it contains.

xMIx: High-Performance Serving-Time Platform for Mechanistic Interpretability Apps

Mechanistic interpretability (MI) has emerged as a powerful approach for analyzing and intervening in inference computations, with a growing number of applications such as jailbreak attempt detection, truthfulness evaluation, and hallucination detection. Unfortunately, MI deployment in production model-serving systems is currently not practical, as most existing MI frameworks introduce prohibitively high runtime overheads. The fundamental problem is that MI functions do not compose cleanly with served models: they fragment deployment, often force draining requests and rebuilding serving state, and conflict with critical performance optimizations such as continuous batching and CUDA-graph execution, essential for production deployments. We present xMIx, a serving-native framework for deploying MI applications in production inference serving environments. xMIx enables attaching MI functions to a predefined set of locations in the model runtime, interposing on activations within the layers and residual streams. xMIx supports conditional invocation of MI functions depending on the outputs in preceding model layers. Multiple MI applications can be deployed in a single model instance. xMIx compiles them all into the serving path but activates them dynamically at runtime only when necessary, with negligible performance cost, and without requiring a separate model instance or alternative execution stack. We integrate xMIx with the vLLM serving system and evaluate it across three major models and seven diverse MI applications. xMIx achieves performance comparable to native vLLM execution, incurring a slowdown of 1.3% mean inter-token latency (ITL), 1.2% for tail P99 ITL, 2.6% for mean time to first token (TTFT), and 1.6% for mean total token throughput (TTT).

An Agentic Orchestration of Atomistic Simulations

Atomistic simulations are central to materials design, but their execution involves complex, multi-step workflows that require significant human expertise. Here, we present an agent-based system embedded within the URSA (Universal Research and Scientific Agent) framework that automates the design, execution, and validation of atomistic simulations, demonstrated using the Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) tool. Our system autonomously selects interatomic potentials, constructs and runs simulations, and performs iterative error recovery within a closed-loop workflow. We evaluate the scientific reliability of the agent by benchmarking its outputs against LAVA, a high-throughput toolkit for LAMMPS and the Vienna Ab initio Simulation Package (VASP) calculations. Our framework reduces manual intervention and trial-and-error, thereby improving the rigor, reproducibility, and scalability of atomistic modeling.

HyCE-RAG: Hypergraph Chain-of-Evidence Retrieval-Augmented Generation for Explainable Multi-hop Question Answering

Multi-hop question answering requires systems to retrieve evidence from multiple documents and connect scattered facts into a coherent reasoning process. Standard retrieval-augmented generation (RAG) mainly relies on semantic similarity between a query and text chunks, and therefore often fails to model structural relations among entities, facts, and evidence units. Graph-based RAG improves this by introducing graph-structured knowledge, but pairwise edges are still limited in representing higher-order associations involving multiple entities and contexts. We propose HyCE-RAG, a Hypergraph Chain-of-Evidence Retrieval-Augmented Generation framework for explainable multi-hop question answering. HyCE-RAG organizes entities, relations, and contextual evidence into hyperedges, builds a query-aware evidence hypergraph, and performs confidence propagation over entity--hyperedge incidence structures. It then uses confidence-guided evidence assembly to select, connect, and rank evidence paths before answer generation. The scoring process jointly considers semantic relevance, entity connectivity, evidence coverage, relation reliability, extraction confidence, and propagated confidence. By providing the language model with structured evidence chains rather than flat retrieved passages, HyCE-RAG supports more faithful and interpretable reasoning. Experiments on HotpotQA, 2WikiMultihopQA, MuSiQue, and two GraphRAG-Bench subsets show that HyCE-RAG consistently outperforms standard RAG and graph-based RAG baselines in answer accuracy, context relevance, and faithfulness. These results suggest that hypergraph-based evidence organization is a promising direction for post-retrieval reasoning in complex question answering.

Differencing the Diffusion Trajectory toward Uncertain Components for Time Series Forecasting

Diffusion models have become a widely used framework for probabilistic time series forecasting, modeling the distribution of future values given an observed history. In time series forecasting, however, the future continues the observed history, creating an asymmetry the standard diffusion process leaves unaddressed, with slowly-varying content largely determined by the observed continuity while higher-frequency dynamics carry most of the residual uncertainty. Existing diffusion-based forecasters decouple this asymmetry through an external rule before generation, leaving the corruption trajectory blind to which parts of the target the history can already anchor. We propose DiffDiff, a diffusion framework that embeds this predictability asymmetry into the diffusion trajectory itself, so that a single end-to-end diffusion process becomes aware of which parts of the target the history can already anchor. DiffDiff makes the forward operator step-dependent so that the noisy intermediate state progressively shifts from the target itself toward its second-order differenced structure, while a conditioning pathway supplies the denoiser with both value-domain and differential history information balanced by a stage-adaptive gate at each diffusion step. The terminal distribution approaches a standard Gaussian, preserving compatibility with existing samplers. On seven benchmarks across four prediction horizons, DiffDiff outperforms six diffusion baselines, and our analysis confirms that DiffDiff concentrates the diffusion's generative effort on the most uncertain components of the target while relieving it from rebuilding the history-anchored content.

Chart Deception in Vision-Language Models: From Vulnerability to Mitigation

Information visualizations are widely used to communicate patterns, trends, and outliers, yet deceptive design choices-such as truncated or inverted axes, distorted aspect ratios, inappropriate encodings, and misleading color mappings-can systematically alter interpretation while preserving the underlying data. As Vision-Language Models (VLMs) are increasingly used for chart understanding and analytical reasoning, assessing their robustness to such deceptive visualizations has become critical for trustworthy data analysis. We introduce VisDeception, the first controlled paired benchmark for evaluating the robustness of VLMs to misleading chart designs. The benchmark contains 1,600 paired faithful and misleading charts spanning eight major categories of deceptive visualization tactics, where each misleading chart is paired with a faithful counterpart generated from the same underlying data. To isolate deception-induced reasoning errors from baseline chart-understanding errors, we introduce the Deception Score, a paired evaluation metric that quantifies how misleading visualizations shift model responses away from the faithful interpretation of the data. Across 32,000 responses from 10 state-of-the-art VLMs, we find that even advanced models remain highly vulnerable to deceptive visual manipulations. To improve robustness, we further propose an inference-time multi-agent mitigation framework that grounds reasoning in structured chart metadata extracted from the visualization before answer generation, enabling models to reduce the influence of deceptive visual cues without requiring explicit user instructions. Together, our findings reveal important reliability gaps in current chart-understanding systems and establish benchmark-driven evaluation, deception-aware metrics, and structured reasoning as promising directions for developing more trustworthy VLMs for visual analytics.

DeepLook: Deeper Thinking with Lookahead

Inference-time scaling has emerged as a powerful paradigm for improving large language model reasoning, often delivering larger gains on difficult reasoning tasks than parameter scaling alone. However, existing approaches remain inefficient in how compute is allocated within a reasoning trace. Motivated by the observation that reasoning failures often exhibit an early onset of uncertainty before a wrong answer become explicit, we introduce DeepLook, a training-free monitor-and-intervene decoding framework that concentrates lookahead compute at uncertainty bottlenecks. DeepLook aggregates token-level confidence into segment-level signals, triggers when confidence drops relative to recent history, and explores candidate continuations with fixed-horizon lookahead. Branches are ranked by Average Lookahead Confidence (ALC), the average segment-level confidence over rollout continuations, then pruned and aggregated through voting. On four competition-style mathematics benchmarks across DeepSeek-R1-8B, Qwen3-32B, GPT-OSS-20B, and GPT-OSS-120B, DeepLook shifts the accuracy--token-cost Pareto frontier: it improves accuracy over DeepConf-low in 11 of 16 settings while reducing dataset-level token generation by 87.3% on average, including gains of +3.1 on AIME25 with Qwen3-32B and +8.8 on BRUMO25 with GPT-OSS-20B. These results show that selective, future-aware intervention yields substantially stronger accuracy--cost trade-offs than uniformly scaling complete reasoning trajectories. Code is available here.

Group Preference Collapse in Personalized Multimodal Large Language Models

Personalized multimodal large language models (MLLMs) aim to generate user-specific responses, but existing methods mainly rely on profile-level information and overlook diverse user preferences. We identify group preference collapse, where multi-user personalized MLLMs become insensitive to individual preferences and drift toward dominant population-level choices due to suppressed preference signals and unreliable preference use during generation. We propose PrefMoE, a preference-centric framework that separates stable profile information from preference-related representations. PrefMoE decomposes preferences into shared prototypes and personalized residuals, preserves individualized residuals with imbalance-aware learning, counterfactual pseudo-user augmentation, and residual decorrelation, and routes profile and preference factors through separate LoRA adaptation paths. Experiments across multiple MLLM backbones show that PrefMoE improves preference-sensitive personalization while substantially reducing preference collapse. Project page: https://prefmoe.github.io/.

Evaluating LLMs as Interpretable Controllers for Dynamical Systems

Large Language Models (LLMs) are increasingly used for decision-making and reasoning tasks, yet their potential as controllers for physical systems remains largely unexplored. This work investigates whether LLMs can function as interpretable controllers for a dynamic thermal environment, examining their ability to follow setpoints, interpret natural-language commands, reason about actuator effects, and incorporate prior model-based knowledge. Five LLMs of varying scales are evaluated under multiple scenarios, including settings with penalties on heater or fan usage and cases where the models have access to a physics-based prediction tool. The results show that control performance depends on model complexity: while low- and mid-scale models frequently misinterpret actuator dynamics or generate inconsistent reasoning, high-complexity models such as Qwen-3~14B and GPT-4o achieve accurate temperature tracking, stable actuator usage, and coherent explanations aligned with physical principles. Incorporating a physics-based model significantly improves control smoothness and energy efficiency by enabling anticipatory decision-making. A detailed reasoning taxonomy further reveals a clear progression from causal misinterpretation in smaller models to cohesive and temporally aware reasoning in larger ones. The findings demonstrate that LLMs can act as interpretable controllers when sufficiently capable and appropriately grounded in domain knowledge, highlighting promising opportunities for hybrid model-based and language-driven control strategies that can provide plausible explanations.

Tokengeist: Multi-Turn Attribution Tracing in Agentic Conversations

When a language model produces a response in a multi-turn conversation, which tokens from prior turns shaped that answer, and how did those dependencies propagate across prior turns? Existing context attribution methods process the full context in a single pass, recovering surface-level dependencies but missing the layered, non-linear structure of real-world dialogues and multi-step reasoning tasks. We introduce multi-turn context attribution (MTCA): given a target span in a model response, the task of tracing attribution backward across turns to identify not only which prior turns were directly relevant, but also how those turns themselves depended on earlier context. We propose Tokengeist, an attribution-method-agnostic and scalable framework that recovers full dependency paths by casting attribution as a recursive traversal of a directed acyclic graph (DAG) over conversation turns. We will release MTCABench, a benchmark of 3,845 target spans across 665 multi-turn conversations, annotated with gold provenance graphs reaching depths of up to 14, across four dependency types. Across four open-weight models, flat attribution methods fail to recover multi-hop dependencies, achieving under 20% source recall, while Tokengeist reaches 90%. Our results reveal systematic failure modes of single-pass attribution -- which we term provenance collapse -- and motivate attribution methods that reason recursively across turns.

Decentralized Granular Access Control for Agentic AI Systems in Critical Infrastructure

The deployment of autonomous AI agents in production infrastructure introduces fundamental security challenges that traditional role-based access control (RBAC) models cannot address. Unlike deterministic automation, AI agents exhibit stochastic behavior, making conventional trust models insufficient for governing their access to critical systems. This paper presents a decentralized, multi-layered access control architecture designed specifically for agentic AI systems operating in critical cloud infrastructure. Our framework introduces four key innovations: (1) a compound identity model that binds agent actions to delegated human authority, (2) a hierarchical permission system spanning five granularity levels from global platform access to per-parameter constraints, (3) a decentralized policy ownership model where tool teams independently govern their authorization boundaries, and (4) progressive trust escalation with safety interlocks that prevent autonomous agents from executing high-risk operations. We ground our design in the OWASP Top 10 for LLM Applications (2025) threat taxonomy and demonstrate how each architectural decision mitigates specific attack vectors. Deployed in production at a major cloud provider managing network infrastructure across hundreds of datacenters, the system enforces granular access control for 20+ specialized AI agents and 60+ deterministic playbooks processing thousands of operations daily while maintaining zero unauthorized write operations over eight months of production deployment. We present empirical data on access pattern distributions, denial rates, and the effectiveness of layered authorization in preventing privilege escalation by non-deterministic actors.

DynaResize: Runtime GPU Reallocation for Disaggregated LLM Post-Training

RL-based LLM post-training increasingly disaggregates Rollout and Training across separate GPU resources, but static GPU partitioning suffers from severe pipeline bubbles under long-tail rollout latency. We present DynaResize, a runtime GPU reallocation system that dynamically switches GPUs between Rollout and Training to balance stage execution times without changing RL semantics. DynaResize decomposes resizing into fine-grained operations and removes non-startup-critical work from the critical path through communicator reuse, bounded state staging, and hysteresis-based resizing. Experimental results show that DynaResize can improve end-to-end throughput by 66.5% and reduce total execution time by 33% over the optimal static configuration, while hiding 27% of role-switching overhead.

Opti-Q: A Constraint-Based Optimization Framework for Multi-LLM Question Planning

While large language models (LLMs) enable strong question answering (QA), budgeted deployment is complicated by nondeterminism and heterogeneous resource profiles (cost, latency, and energy). We present OPTI-Q, a database-inspired, cost-based optimizer that implements a plan-before-execute paradigm for multi-LLM orchestration. OPTI-Q models LLM invocations as physical operators in an execution DAG and, for each question, searches for plans that optimize answer quality (QoA) while trading off financial cost, latency, and energy under user-specified resource constraints. Plans can include sequential operators that pass intermediate answers as context and parallel/blend operators that run models concurrently and merge their outputs. To search this space without executing each candidate plan, OPTI-Q uses PERFDB, a statistics catalog populated and refreshed from benchmarks and execution traces, to estimate the QoA and resource costs of both individual operators and composed subplans. Using these estimates, OPTI-Q performs Pareto-frontier search and selects a final plan based on user preferences. On MMLU-Pro and SimpleQA under user-specified budgets, OPTI-Q improves average QoA by ~58% and ~41% over baselines at comparable cost, demonstrating that database-style planning yields better quality-resource trade-offs for multi-LLM QA.

CHS-SQL: A Text-to-SQL approach based on Confidence-Guided Heuristic Search Schema Linking process

Recently, there have been several works in the Text-to-SQL domain that utilize Small Language Models (SLMs) for training. These approaches achieve performance close to that of large models in generating SQL, using only the computational power of a single NVIDIA RTX 4090 GPU, while also ensuring data security. Most existing methods filter out redundant tables and columns during Schema Linking to improve Text-to-SQL accuracy. However, they do not consider the precision-recall trade-off when selecting the candidate schema subset. Our research found that both the precision and recall of Schema Linking directly affect the final SQL accuracy. Therefore, we propose a novel framework for efficiently fine-tuning SLMs on Text-to-SQL tasks, CHS-SQL, that not only balances precision and recall but also improves overall performance on Text-to-SQL tasks. Its main innovation lies in the Schema Linking phase, where a heuristic search combined with model internal confidence is employed to achieve an optimal precision-recall trade-off. This elaborated mechanism maximizes the precision of relevant schema candidates for the generated SQL queries while suppressing irrelevant noise. The same strategy is further applied during SQL generation to refine candidate queries while helping the SLM to avoid trapping in a local optimum. Our method achieves state-of-the-art (SOTA) results on Text-to-SQL tasks via SLMs.

TokenMem: Faithful Knowledge Injection for Frozen LLMs

Retrieval-augmented generation (RAG) enhances large language models (LLMs) with external knowledge, but suffers from knowledge conflicts: when retrieved information contradicts parametric memory, the shared self-attention pathway produces unpredictable outputs. We present TokenMem, a lightweight memory system that injects knowledge into frozen LLMs through a dedicated cross-attention channel, bypassing competition with parametric memory in the residual stream. TokenMem trains only a thin gating adapter ($\sim$3-7M parameters) via a two-phase curriculum: first learning general knowledge utilization, then strengthening faithful compliance under counterfactual knowledge. In controlled experiments on five models spanning three families (Qwen3-4B/8B/14B, LLaMA-3.1-8B, OLMo-3-7B), TokenMem achieves 69-70% Knowledge Compliance (KC) on counterfactual benchmarks, compared to 20-52% for vanilla RAG, a gap of up to 49 percentage points. Ablation studies show that the two-phase curriculum is critical: removing Phase 2 collapses KC to near-zero. Mechanistic analysis reveals that the gate adapter learns a conflict-aware, layer-specific injection strategy without explicit supervision.

Masked Distillation: Internalizing the Chain-of-Thought in Language Models

Large Reasoning Models (LRMs) produce long, explicit chains of intermediate steps before generating a final answer at inference time. These intermediate traces dominate latency, memory usage, and serving cost, even though the final answer correctness is not causally related to the trace correctness and the trace length is not a reliable indicator of the problem complexity. This raises a natural question: can the computation expressed in these intermediate tokens be internalized into the parameters of a language model, enabling it to produce answers directly (or with much shorter intermediate traces)? We introduce \textit{masked distillation}, a knowledge-distillation framework in which a student LLM is trained to predict only the solution tokens conditioned on the question, while a reasoning teacher provides feedback on the student's responses after conditioning on the question and its own CoT trace. We instantiate this framework in two settings: (i) a \textit{self-distillation} setting, in which the same model serves as the teacher in thinking mode and as the student in non-thinking mode, and (ii) a \textit{dual-model} setting, in which a larger reasoning teacher supervises a separate smaller non-thinking student over the solution tokens. By treating intermediate tokens as a scaffold which reasoning models use to fit over the solution tokens, We additionally vary the length of intermediate-token scaffolding the student is supervised on, interpolating between full internalization (the student emits only the solution) and no internalization (the student emits the full trace before the answer). We evaluate the framework through controlled experiments on two reasoning domains: GSM8K (grade-school arithmetic) and Countdown (a number-puzzle search task).

VlogReward: Learning Multi-Dimensional Evaluation for Vlog Editing

The rapid rise of vlogs as a personalized storytelling medium has created a demand for automated systems to evaluate and refine vlog editing plans. However, vlog assessment is highly subjective and remains challenging due to a lack of standardized criteria, dataset and benchmark, and effective reward models. To address these challenges, we define a comprehensive vlog evaluation framework guided by professional vlog creators and product managers, establishing a taxonomy of six key dimensions, i.e., Creativity, Consistency, Concept Design, Cinematography, Narration, and Pacing. Subsequently, we curate a large-scale dataset of 100k vlog edits and a dedicated benchmark, VRMBench, to evaluate the vlog rewarding capabilities of Multimodal Large Language Models (MLLMs). Finally, we present VlogReward, a robust vlog reward model that can provide both fine-grained multi-dimensional scores and actionable feedback for iterative refinement. Technically, we enhance the Group Relative Policy Optimization (GRPO) framework by introducing an adjustable inter-group comparison reward, which mitigates the "direction blindness" issue of standard GRPO and enables the model to better distinguish varied-quality edits. VlogReward achieves state-of-the-art results that significantly outperform existing MLLMs, including GPT-5 and Gemini-3-Pro. We hope that our study can help vlog creators and foster automated vlog evaluation and refinement systems.

Evolving from Lessons: Skill-Augmented Table Graph Reasoning for Operation-wise Table Question Answering

Table Question Answering (TableQA) aims to reason over tables to answer user queries. Existing research treats all questions uniformly and evaluates solely through overall accuracy, obscuring a critical reality that LLMs excel at simple lookups yet struggle with complex operations like aggregation and arithmetic. To reveal this disparity, we introduce a novel \emph{Operation-wise TableQA} task with a fine-grained question taxonomy and release two datasets named WikiTQ-ow and TabFact-ow for evaluation. As for modeling bottlenecks, existing methods flatten tables into linearized texts, disrupting inherent structures and inducing the ``lost-in-the-middle'' issue, which poses a primary barrier to complex cross-row reasoning. Moreover, they typically reason from scratch, neglecting reusable patterns shared across similar operations. To address these limitations, we propose a Skill-augmented Table Graph Reasoning (SkillTGR) framework for self-evolving structured reasoning. Specifically, SkillTGR represents tables as attributed graphs with explicit row-column-cell structures, where LLMs plan and execute dynamic chains to retrieve evidence subgraphs for graph traversal reasoning. Based on this, SkillTGR builds a hierarchical SkillBank to distill reason trajectories into abstract skills under cognitive heuristics, then hybrid retrieves both successful and failed skills for contrastive augmented table graph reasoning, thereby enabling the continual self-evolution. Extensive experiments demonstrate that SkillTGR achieves superior performance with an average of 5.91\% overall and 6.03\% operation-wise improvement, also reducing 19.76\% token consumption and 27.64\% inference latency. Our codes and data will be released upon publication.

PRESTO: Prefix-Aligned Tree Drafting for Diffusion Speculative Decoding

Diffusion Large Language Models (dLLMs) have emerged as a promising alternative to autoregressive (AR) LLMs, generating tokens in parallel. This makes them effective draft models for speculative decoding (SD), producing an entire block of draft tokens in a single forward pass. Yet existing diffusion-based drafting methods rely on linear drafting, even though dLLMs emit multiple candidate tokens across positions, inducing a large combinatorial space of decoding paths. Consequently, they limit acceptance length and decoding efficiency. To exploit this multi-candidate structure, we apply tree-based drafting to diffusion drafters, enabling exploration of diverse candidate paths. However, we find that naive tree drafting is suboptimal: diffusion marginals are prefix-blind, mismatching the prefix-based AR verification and yielding unreliable path ranking. We propose PRESTO, a principled framework that extends tree-based drafting to diffusion drafters while resolving the fundamental mismatch between diffusion draft confidence and prefix-based AR verification through PREfix-aligned Scoring and priority-based Tree search for diffusion speculative decOding. The key principles behind PRESTO are that (1) candidate ranking should align with the prefix-based nature of AR verification, and (2) tree construction should prioritize candidate paths with high verification potential to maximize acceptance length. Extensive experiments show that PRESTO achieves up to an average of $1.5\times$ end-to-end throughput speedup on the state-of-the-art dedicated diffusion drafter SD and an average of $1.12\times$ on self-speculative diffusion LLMs across diverse benchmarks.

CallBench: A Benchmark for Dual-Goal Coordination in Phone Call Assistants

Target-oriented dialogue systems have demonstrated strong capabilities in completing user goals through interactive conversations. However, existing studies are primarily designed for single, explicit goal completion, while phone call assistants face a proxy setting that requires coordinating the device owner's explicit preset goal with the caller's implicit and dynamic goal. We introduce \textsc{CallBench}, a Chinese benchmark for evaluating dual-goal coordination in phone call assistants. \textsc{CallBench} contains 50,000 complete multi-turn phone call dialogues across six scenarios: takeout, delivery, taxi, work, life, and harassment. It covers regular presets, emergent presets, and no-preset cases, and includes diverse relations between owner-side and caller-side goals, such as alignment, complementarity, irrelevance, and conflict. We further design a preset-aware turn-level evaluation protocol covering semantic understanding, context use, active guidance, response quality, preset compliance, dialogue rhythm, and safety. Experiments on representative dialogue methods show that existing approaches still struggle with this task, highlighting the need for phone call assistants that can make reliable turn-level decisions between two independent goals under proxy constraints.

Answering Path Queries under Linear and Guarded Existential Rules

Ontology-mediated query answering is concerned with the problem of answering queries over knowledge bases consisting of a database instance and an ontology. While most work in the area focuses on conjunctive queries (CQs), navigational queries have gained increasing attention. In this paper, we investigate the complexity of answering two-way (conjunctive) regular path queries ((C)RPQs) over knowledge bases whose ontology is given by a set of guarded existential rules. We first consider the subclass of linear existential rules and show that (C)RPQ answering is NL-complete in data complexity, which matches the data complexity of answering RPQs over plain graph databases (i.e., without an ontology). In combined complexity, both tasks are ExpTime-complete in the general case, but RPQ and CRPQ answering drop to PTime-complete and PSpace-complete respectively if there is a bound on predicate arity. For guarded rules, we provide a non-trivial reduction to the linear case, which allows us to show that the complexity of (C)RPQ answering is the same as for CQs, namely 2ExpTime-complete in combined complexity (ExpTime-complete in the bounded-arity case) and PTime-complete in data complexity.

Fast Cross-Scenario Adaptation of CSI Models via Channel Conditional Parameter Generation

Deep learning has shown strong potential for massive multiple-input multiple-output (Massive MIMO) physical-layer tasks, including channel state information (CSI) feedback and channel estimation. However, environmental heterogeneity can severely degrade CSI models in unseen scenarios, while conventional adaptation requires target-domain data and substantial computation. This paper proposes Channel Conditional Parameter Generation (CCPG), an end-to-end pipeline for rapid deployment of CSI models in dynamic wireless environments. CCPG identifies scene-sensitive adaptation bottlenecks through component-freezing experiments and generates only lightweight LoRA weights instead of full model parameters. It compresses high-dimensional channel features into compact latent conditions using cascaded SVD and a Perceiver Resampler. An energy-based canonicalization mechanism mitigates permutation and sign ambiguities in LoRA weights, while a diffusion-based generator incorporates structural information and an asymmetric size-aware loss for topology-aware parameter generation. Experiments on DeepMIMO and WAIR-D for CSI feedback and channel estimation show that CCPG adapts to new scenarios in about 3 seconds with a single forward pass, without target-scenario training or fine-tuning, and achieves cross-domain recovery performance comparable to costly online adaptation. These results demonstrate that CCPG enables efficient deployment of CSI models in large-scale dynamic wireless scenarios for intelligent 6G communications.

TRACE: Business Rule-Grounded Reasoning Curriculum for Knowledge-Preserving Parametric Tool Retrieval in Enterprise LLMs

Parametric retrieval enables LLMs to retrieve tools implicitly by assigning each API a unique virtual token and training the model to generate it via constrained beam search. Toolsense shows that this regime has two critical drawbacks: it destroys parametric tool knowledge during training, and its beam-search decoding is too slow for real-time deployment. We introduce TRACE (Tool Retrieval via Augmented Chain-of-thought and Enterprise rules), a two-stage curriculum that resolves this dissociation. Stage 1 reuses the multi-format memorization SFT from ToolSense to seed tool knowledge with LoRA. Stage 2 is our core contribution: the model is trained to emit a thinking trace before producing a JSON list of tool tokens, using two data sources -- RRB pairs from ToolSense and queries synthesized to target business rules curated by domain experts -- both augmented with reasoning traces. This training objective preserves Stage 1 MCQ and QA probing accuracy while enabling single-beam greedy decoding at production latency. Evaluated on a combined enterprise catalog of 8,300+ tools across two enterprise product lines, TRACE training for Stage 2 not only preserves but improves tool understanding: MCQ accuracy gains +3.2 pp and QA probing gains +9 pp over Stage 1. On retrieval, TRACE achieves ~86% recall on Domain A and ~60% on Domain B -- compared to embedding baseline performance of ~27% & ~52% -- both with single-beam greedy decoding, making it directly deployable at production latency.

CRAFT: Learn the Schema, Execute the Plan

Enterprise coding agents translate natural-language analytical requests into executable code over proprietary APIs, schemas, and metric definitions. Yet the prevailing deployment pattern injecting exhaustive schema and tool documentation into each prompt increases inference overhead, complicates schema evolution, and undermines reliability in multi-turn analysis. We investigate whether stable schema knowledge and tool-use behavior can instead be acquired through post-training while preserving the consistency required for production-facing analytics. We present CRAFT, a two-stage post-training recipe for schema-grounded coding agents. First, schema-stripped PLAN supervised fine-tuning learns domain-structured plans and executable behaviors from validated trajectories without exhaustive prompt-time schema injection. Second, execution-shaped reinforcement learning aligns the policy for tool selection, code quality, plan-code consistency, and recovery from failed executions. Training trajectories are curated through a Tri-Gate filter combining execution validation, data-integrity checks, and LLM-judge reasoning audit. We evaluate CRAFT for planned rollout in advertising analytics, covering campaign performance analysis, metric drill-downs, entity-level performance analysis, and multi-turn analytical refinement. The enterprise evaluation environment incorporates beta APIs as the agent-facing tool surface and spans 25 schema-linked core entities and 30 agentic workflows. Relative to a schema-stuffed baseline, CRAFT improves composite Agent Score by +9.6 pp, consistency by +4.1 pp, and multi-turn coherence by +4.2 pp, while reducing input-token burden by approximately 9x and schema-discovery loops by up to 5x. We further report deployment tradeoffs, reward-shaping limitations, and training-infrastructure extensions required for multi-turn tool-use reinforcement learning in enterprise settings.

Reason Before You Retrieve: Agentic Planning for Multi-modal RAG

Multimodal retrieval-augmented generation (mRAG) aims to answer image-text queries with external knowledge, but most existing systems still retrieve directly from raw multimodal input over a flat evidence space. This design often struggles with two key challenges: the retrieval target is under-specified because the question intent must be grounded to the correct visual referent, and the search space is weakly structured, forcing semantically distinct evidence to compete in a single global ranking step. We propose MM-R2, a multimodal agentic retrieval framework that reasons before retrieval by explicitly modeling both what to retrieve and where to search. MM-R2 first constructs an intent-grounded retrieval state from the image-question pair, capturing the information need, grounded referent, and retrieval constraints. It then performs retrieval over a structured KnowledgeMap, where the agent selects relevant retrieval units before issuing grounded queries within them. To enable this capability, we build MM-R2-Traj, a large-scale trajectory dataset of multi-step retrieval processes, and adopt a two-stage post-training strategy with supervised fine-tuning and GRPO. Experiments on Infoseek and Encyclopedic VQA datasets show that MM-R2 substantially outperforms strong baselines on answer accuracy while also yielding more interpretable and verifiable retrieval trajectories.

DocHRL: A Hierarchical Reinforcement Learning Framework for Cost-Optimised Document Classification

Real-world document classification pipelines typically apply the same sequence of models to every incoming document, regardless of its complexity or type. This leads to inefficient use of compute and human resources: simple documents are over-processed while difficult ones may not receive enough scrutiny. We introduce DocHRL, a hierarchical reinforcement learning framework that learns to adaptively and dynamically select the most cost-effective classification policy on a per-document basis. DocHRL formulates document classification as a sequential decision problem with a two-level policy hierarchy: a top-level policy selects among broad options (vision classifiers, LLMs, OCR, and human-in-the-loop review), while option-specific sub-policies choose the concrete model or tool to invoke. The reward signal is the negative total expected cost, which captures inference cost, cost of misclassification, and cost of human labelling. Trained with Proximal Policy Optimisation on the RVL-CDIP benchmark, DocHRL achieves a macro F1 of 0.973 across 16 document classes while reducing average per-document cost to 2.74 normalised units compared to substantially higher costs incurred by fixed standalone classifiers. Our results demonstrate that cost-aware reinforcement learning can simultaneously improve classification performance and operational efficiency in document understanding systems.

Extracting Algorithms in Pre-trained LLMs: A Case on Hidden Markov Models

Large language models (LLMs) display a striking ability to predict next observations from Hidden Markov Models (HMMs) via in-context learning (ICL), but the algorithm underlying this capability remains undetermined: prior work has proposed several candidates without consensus, and none has been grounded in the model's internal activations. We close this gap with a three-stage pipeline. First, we empirically compare LLM behavior against a suite of candidate algorithms and narrow the space to three classes -- though no single class explains LLM behavior across all HMM settings and sequence lengths. Second, we derive theoretical connections between the three classes and show how each can be implemented in-context by a Transformer, validating the construction in a small trained Transformer. Third, returning to pre-trained LLMs, we introduce the Principal Activations Probe (PAP), a layer-wise probing and intervention method that isolates algorithmic signals in model activations. PAP reveals low-dimensional linear representations that causally drive model predictions and track empirical ICL performance. PAP further reveals how these representations shift with properties of the underlying HMM regime; distinct computational stages are localized to different layers. Together, our results connect the in-context behavior of pre-trained LLMs to the underlying internal mechanisms and advance our understanding of how LLMs perform ICL on HMMs.

PTStore (Prefix Tensor Store): Distributed Prefix Caching and Replication for High Throughput Inference Serving

Inspired by the design of client caching in Content Delivery Networks (CDNs), PTStore distributes and replicates popular tensors that form reusable KV cache prefixes, which are the main technique used by state of art approaches to accelerate inferences. This reduces the latency of accessing the KV cache and alleviates load imbalance caused by a disproportionately large number of requests on servers containing popular tensors. Furthermore, thanks to decentralization, PTStore allows the expansion of the size of the KV cache for LLM inference by orders of magnitude. As a result, PTStore can execute inferences on long passage Q\&A datasets 5-6 times more efficiently than current baselines, which do not aggregate memory across different nodes and GPUs and therefore require regenerating the KV cache.

STAIF: A Stage-wise Optimization for Complex Instruction Following

Following complex instructions with multiple explicit constraints remains a fundamental challenge for large language models (LLMs). Existing alignment methods, such as DPO, optimize holistic reward signals that often underemphasize strict satisfaction of individual constraints, particularly under out-of-distribution or multi-constraint settings. In this paper, we propose STAIF, a stage-wise optimization framework that decouples the alignment of subjective (soft) constraints from the optimization of objectively verifiable (hard) constraints. Stage 1 applies preference optimization with multiple negative samples to sharpen sensitivity to soft constraints, while Stage 2 applies Reinforcement Learning with Verifiable Rewards (RLVR) to enforce strict compliance with hard constraints. To support this method, we construct STAINSTRUCT, a high-quality bilingual (English, Chinese) dataset of approximately 31,000 complex multi-constraint instructions. Extensive analyses validate the design of STAIF and show state-of-the-art performance on representative benchmarks against strong baselines, as well as genuine generalization.

ARdena: Scenario-driven control of real-time LLM agents

Large language models (LLMs) have enabled increasingly capable conversational agents, but reliably controlling their behavior in real-time interactive environments remains a significant challenge. Existing approaches often rely on model fine-tuning or alignment procedures that are difficult to adapt to changing interaction requirements. This paper introduces layered scenario-driven LLM control, a framework that enables runtime behavior control through structured prompting. By combining persistent context with scenario-specific constraints, the approach allows agent behavior to be modified during interaction without changing the underlying model. The framework is implemented in ARDena, a real-time multimodal embodied agent that integrates speech interaction, visual perception, tool use, and avatar-based response generation. The proposed approach is evaluated with respect to control effectiveness, response latency, and operational stability. The results demonstrate that scenario definitions alone can produce substantially different interaction behaviors while maintaining stable real-time operation, highlighting the effectiveness of scenario-driven prompting for controlling LLM agents.

KG2Code: Bridging Knowledge Graphs and Large Language Models via Executable Code for Question Answering

Recent research has explored the integration of knowledge graphs (KGs) with large language models (LLMs) to enhance their performance on downstream knowledge-intensive tasks, particularly knowledge graph question answering (KGQA). Existing approaches primarily combine LLMs with KGs through retrieval-augmented generation (RAG)-based, agent-based, and SPARQL-based methods. Although these methods have achieved notable success, they still suffer from several limitations, including structural information loss, unfaithful reasoning, and limited flexibility and generalization. To address these challenges, this paper proposes KG2Code, a novel approach that transforms knowledge graphs into a code-based representation, preserving structural semantics while naturally aligning with the code-aware pretraining of modern LLMs. Based on KG2Code, KG2Code-QA is further introduced as a KGQA framework that formulates KGQA as a code generation task. This formulation enables the generation of verifiable reasoning traces and executable code, thereby substantially mitigating the impact of hallucinations. In addition, an automated pipeline is developed to construct a large-scale, high-quality code corpus for effectively training open-source LLMs on KG2Code-QA. After training, LLMs are able to perform KGQA in zero-shot scenarios. Extensive experiments demonstrate that the proposed approach significantly outperforms existing KG-enhanced LLM methods for KGQA, while exhibiting strong generalization to unseen KGs. The code and data are available at Github.

Do Language Models Converge to Themselves? Recursive Self-Refinement as Textual Relaxation

Large language models are increasingly used in recursive refinement workflows, where an initial draft is repeatedly revised by the same model. Despite their growing use, the long-term dynamics of such workflows remain poorly understood. Does repeated refinement continue to improve outputs indefinitely, or does it converge toward a stable textual form? We study recursive self-refinement as a dynamical process in which repeated LLM revision drives text toward a model-preferred soft fixed-point region. Using GPT-5.5, we generate 10-step refinement trajectories for 50 ICML 2025 abstracts under both default-temperature and deterministic decoding, and additionally evaluate 15 ICML 2020 abstracts. We analyze normalized edit distance, exact and approximate fixed points, word-count stability, exponential relaxation, and external LLM-as-a-judge evaluation. Across all settings, refinement trajectories rapidly saturate. Most edits occur within the first few iterations, after which trajectories enter a soft fixed-point region with only minor surface-level changes. Deterministic decoding reaches exact fixed points earlier and exhibits smaller residual fluctuations than default-temperature decoding, while both achieve universal approximate convergence. The average edit magnitude follows a consistent exponential relaxation pattern, suggesting convergence toward a model-preferred textual equilibrium rather than open-ended optimization. External evaluation indicates that converged abstracts improve clarity, conciseness, and scientific style while preserving technical meaning. These findings support a dynamical-systems view of LLM self-refinement and motivate practical stopping criteria based on edit-magnitude saturation.

MINT-V2X: A Mobility-Integrated Network Trajectory Dataset for Predictive Resource Management

Vehicle-to-Everything (V2X) communication systems are based on datasets that not only contain vehicle trajectory data but also wireless network parameters with a realistic level of fidelity, enabling the creation of prediction and optimization models. There is a very critical research infrastructure gap today, and publicly available datasets are likely to be limited to one of the two: mobility or network parameters, and rarely provide a single, integrated view that combines both. This paper introduces MINT-V2X, a comprehensive dataset generated by coupling SUMO traffic dynamics with OMNeT++/Simu5G network simulation. The validation framework is composed of 14 standardized tests based on 3GPP Release 14 (C-V2X), ETSI standards and Shannon capacity theory. The resulting dataset contains 9.87 million synchronized data points from 1,386 vehicles from 15 roadside units (RSUs) during 3 hours of urban traffic simulation. We demonstrate strict algorithmic consistency through network metric correlations (CQI-SINR: 0.993; SINR-PDR: 0.946). Finally, we demonstrate the value of the dataset by conducting an RSU load prediction case study, showing that using trajectory data yields better predictive performance than network-history-only baselines. The dataset, experiments, and complete SUMO configuration files are available in the GitHub repository to facilitate reproduction on alternative simulation stacks.

EventOD: Event-Aware OD Flow Generation via LLM-Guided Semantic Modulation

Estimating origin-destination (OD) flows under disruptive events is important for disaster response and urban resilience. Existing deep OD models trained on routine mobility often degrade when extreme events abruptly alter regional functions and population activities, while retraining a new generator for each event is impractical under limited event-time supervision. We propose EventOD, an event-adaptive OD generation framework that steers a pretrained OD generator using structured event semantics. EventOD first uses a large language model to infer region-level functional and demographic control vectors from coarse event observations. It then learns two lightweight adaptation modules, AlphaNet and BetaNet, to calibrate the magnitude of these semantic shifts, and further introduces a retrieval-augmented fallback pathway for scenarios with sparse supervision. The resulting event-conditioned features are injected into a pretrained graph diffusion OD model through input-level modulation, enabling event-aware adaptation without updating generator parameters. Experiments on hurricane- and pandemic-induced mobility across U.S. counties show that EventOD consistently improves both reconstruction accuracy and distributional fidelity over strong baselines. Source code is available at https://anonymous.4open.science/r/EventOD-5C11/.

StanceBench: A Benchmark for Audio LLM-Based Interpersonal Stance Evaluation from Speech

Speech-to-speech dialogue models increasingly depend on prosody and interactional nuance to convey social intent, yet benchmarks for these cues remain limited. We introduce StanceBench, a benchmark for measuring interpersonal stance in conversational speech and evaluating audio-capable LLMs as automated judges. Using the Seamless Interaction corpus, StanceBench (1) specifies 9 stance dimensions via role-prompt poles, (2) standardizes single-speaker and interaction-based evaluations, and (3) reports LLM-as-a-judge robustness, bias, and stance inference. Across evaluated stances, empathy and politeness are the easiest. Warmth and assertiveness are moderately separable with positivity skew/asymmetry. Honesty is the hardest and shows high prompt order bias, consistent with needing cross-turn evidence. Attentiveness is separable but aligns weakly with humans. Interaction stances are more context-sensitive, with threshold gaps and high variance, especially conflict regulation.

TRE: Training-Free Hallucination Detection for Diffusion Language Models

Diffusion large language models (D-LLMs) have recently gained increasing attention, yet their reliability is significantly hindered by the hallucination problem. Existing hallucination detection approaches for D-LLMs mainly follow a training-based paradigm, relying on data-driven training to optimize the detector. Such reliance not only limits their generalizability across domains models but also incurs additional training cost and deployment overhead. To address these limitations, we propose TRE, a training-free hallucination detection metric for D-LLMs. TRE is a parameter-free and single-run metric that estimates hallucination risk directly from the entropy signals of a single generation, without requiring any detector training or repeated sampling. TRE extracts entropy signals within the D-LLM decoding process along both the spatial and temporal dimensions. From a token-level spatial perspective, we focus on revealing tokens as the most informative carriers of uncertainty, capturing where uncertainty is actively committed. From a diffusion step-level temporal perspective, we empirically identify the dominance of late-step entropy and hence aggregate these signals with a simple linear weighting scheme to obtain TRE. Extensive experiments on multiple D-LLMs and QA datasets demonstrate that TRE achieves competitive performance, while enjoying strong generalizability, efficiency, and robustness.

CuraWeb: Joint Optimization of Quality, Redundancy, and Diversity for Web-Scale Pretraining Data

Open-web corpora curated via highly selective filters, such as FineWeb-Edu and DCLM, constitute the core of LLM pretraining data and have significantly advanced LLM performance. However, these pipelines typically rely on singular optimization objectives, which inevitably narrows distributional diversity and marginalizes long-tail knowledge, thereby restricting data coverage and underutilizing the vast potential of the open web. To address this limitation, we propose a novel curation paradigm that shifts from linear pruning to the joint optimization of quality, redundancy, and diversity. This framework synergizes dual-track cleaning (rule-based and model-driven) with hybrid deduplication (n-gram and semantic), while employing a multi-objective sampler to balance informational quality with distributional breadth. Applying this framework to Common Crawl, we construct CuraWeb, a 2T-token English corpus. Unlike existing resources, CuraWeb establishes an industrial-grade standard for data curation by recovering a more holistic data distribution with enhanced diversity and minimal redundancy, achieving broader coverage of long-tail knowledge across diverse domains. Experimental evaluations at the 3B scale demonstrate that CuraWeb significantly outperforms state-of-the-art baselines, yielding an average performance gain of 1.8\% across a wide range of benchmarks, particularly in knowledge-intensive and reasoning tasks.

Beyond Block Boundaries: Multi-Block Editing for Diffusion Large Language Models

Block diffusion has emerged as the dominant paradigm for scaling discrete diffusion language models (dLLMs), because decoding text in fixed-size blocks preserves parallel generation within each block while keeping the quadratic attention cost tractable. However, this efficiency comes with a structural limitation: tokens near the end of a block are generated without access to future cross-block context, and once a block is finalized, its uncertain predictions become irreversible context for all subsequent blocks. This creates a block boundary problem, in which uncertainty accumulates toward block boundaries and early mistakes propagate throughout later generation. To address this issue, we propose Multi-Block Editing (MBE), to mitigate this problem by editing decoded tokens based on cross-block context. Following this principle, MBE first proposes a training-free decoding algorithm to edit the decoded tokens in previous blocks, which is achieved by re-opening a full-attention window over selected blocks. Given the mismatched attention mechanism between block diffusion training and MBE inference, MBE further introduces a supervised Fine-tuning strategy, which equips the model with bidirectional attention masks that progressively expands the editing span. Furthermore, it also extends SGLang with a multi-shape CUDA Graph pool and fine-grained KV cache control to make these variable-length editing passes efficient in practice. Experiments on LLaDA2.1-Mini across 13 benchmarks show that training-free MBE outperforms all existing decoding baselines while maintaining comparable throughput, and MBE SFT further brings a performance gain of 2.7. The largest improvements appear on tasks requiring strong long-range consistency, including +13.3 on AIME 2025 and +5.9 on ZebraLogic, demonstrating the effectiveness of MBE.

Obliviate: Efficient Unlearning in Recommender Systems

Machine unlearning is becoming increasingly critical in the context of data privacy regulations, particularly for recommendation systems that are directly trained on user interaction data. The goal of this work is to remove requested interaction data and their downstream influence from trained model while preserving recommendation quality, and to do so without incurring the substantial computational cost of full retraining. Existing approaches exhibit several limitations, including limited unlearning completeness and degradation in recommendation performance, while having substantial computational overhead. In this paper, we propose Obliviate, an efficient two-stage unlearning framework for recommender systems that achieves high unlearning completeness while maintaining good utility. In the first stage, we introduce a Low-Rank Unlearning Adapter (LUA), which employs a lightweight Hessian proxy to enable curvature-aware and efficient unlearning through localized low-rank adapters rather than full parameters. In the second stage, we propose Locality-Aware Calibration (LAC), a lightweight refinement stage that updates only the adapter parameters to improve the performance by enforcing unlearning via ranking-based objectives while preserving utility through knowledge distillation. Extensive empirical evaluations demonstrate that Obliviate achieves high level of forgetting with minimal loss in recommendation quality and at significantly reduced computational cost, offering a practical and scalable solution for large-scale recommender systems.

Reinforcement Learning for Heterogeneous Sensor Selection in Maritime Surveillance

This paper presents an information-gain-guided reinforcement-learning sensor-selection framework for single-vessel tracking in heterogeneous maritime sensor networks. The proposed approach is motivated by information-theoretic sensor management: instead of activating all sensors or repeatedly performing computationally expensive online expected-information-gain evaluation, a learned policy selects one tracking-relevant sensor at each decision epoch. A Bayesian sequential Monte Carlo tracker estimates the vessel state from noisy measurements and provides a belief representation for scheduling under nonlinear and non-Gaussian conditions. A Proximal Policy Optimization agent selects one of five sensors deployed in a georeferenced simulation of the CMMI Smart Marina testbed at Ayia Napa Marina, Cyprus. The agent observes belief-state, detection-history, coverage, sensor-geometry, and realized-information-gain features. The reward is defined as a realized-information-gain term gated by an observability mask. Final-test simulations compare the proposed framework with random single-sensor selection, always-on sensing using all sensors simultaneously, and the expected-information-gain sensor-selection baseline proposed in our previous work. Results show that the learned policy achieves tracking performance close to always-on sensing while activating only one sensor per decision time step and avoiding the computationally expensive online entropy search required by expected-information-gain selection.

AIR-BENCH Live: An Evolving Safety Benchmark for Foundation Models

Foundation-model safety benchmarks capture the AI risks of their time of publication: as models improve and governments pass new AI-safety legislation, their risk taxonomies become incomprehensive and their attack prompts become ineffective. We present AIR-BENCH Live, a self-evolving successor to AIR-BENCH 2024. An automated update pipeline monitors government regulation and classifies new policies against the current four-tier risk taxonomy, either matching them to existing categories or proposing new granular categories. Then, a multi-agent, persona-driven prompt generation algorithm generates realistic, multilingual prompts with minimal human review, leaving room for improvement with modern jail breaking techniques. This algorithm is used to overhaul legacy prompts and generate prompts for new categories. In our current version, the pipeline has expanded the benchmark from 314 to 335 granular risks, with the 21 new categories drawing from 31 truly novel policy clauses across seven jurisdictions. Evaluating 14 recent models, we find a wide safety spread (from 0.17 to 0.89 among the models judged on their own behavior), that the modernized prompts are on average 0.06 points harder than the 2024 set, with the largest drops concentrated among the most compliant models, and that most models are modestly less safe on non-English prompts. By continuously absorbing new regulation and regenerating prompts, AIR-BENCH Live is designed to evolve alongside a fast-moving field.

How LLM Task-Adaptation Reshapes Alignment: A Multi-dimensional Study of Behavioral and Representational Drift

Post-training is a key mechanism for adapting large language models to downstream tasks. While prior work suggests that task adaptation can alter a model's pre-existing alignment, especially its safety behavior, its broader effects across alignment domains remain poorly understood. We address this gap through a systematic evaluation of representative task-adaptation methods, including supervised fine-tuning (SFT), KL-regularized SFT, and reinforcement learning with verifiable rewards (RLVR) across 15 alignment aspects spanning six key domains: safety, factuality, stance stability, social harm, controllability, and instructability. Our results reveal that post-training does not reshape alignment uniformly. RLVR improves task performance while inducing comparatively small, but non-zero, metric-specific shifts, while SFT leads to substantially larger alignment drift across domains. KL regularization mitigates this effect: stronger reference-model anchoring reduces alignment drift from the baseline, although KL-SFT still falls short of RLVR in preserving alignment. Representation-level analysis further supports this pattern, with shifts in alignment-relevant representations tracking behavioral drift. Together, these results show that task adaptation is not merely a capability-improving step, but an alignment intervention in its own right, motivating multi-dimensional alignment evaluation as a standard component of post-training pipelines.

DOSA: A Tree-Guided, Self-Regressive Framework for Long Document Structure Analysis

In visually-rich documents, information is encoded not only in individual page objects such as tables, headers, and text blocks, but also in the structural relations among them, making document structure analysis fundamental to information retrieval and document understanding. However, accurately inferring such relations remains challenging in multi-page documents with long-range dependencies and heterogeneous layouts. To address this, we propose a tree-guided and self-regressive framework, termed DOcument Structure Analyzer (DOSA), for inferring relations among page objects and reconstructing document-level semantic trees. DOSA processes documents chunk-by-chunk, fusing visual, textual, and layout features for each page object and predicting hierarchical and ordering relations. The predicted relations are used to incrementally construct a semantic tree, which is then leveraged as structural context to guide inference on subsequent chunks. Experimental results on five benchmarks demonstrate the effectiveness of DOSA, with improvements of up to 4 F1 points and 19 TEDS points on DocHieNet, the most challenging multi-page hierarchy benchmark.

A Vocabulary for Multi-Agent Automated Research Systems

We introduce a vocabulary for automated research systems built from one or more agents to make their design choices easier to describe and compare. The vocabulary specifies 1) who the agents are, 2) what operations are available in the system, 3) who may invoke them, 4) how agents communicate, 5) what information is visible within and across runs, 6) how the next action is chosen, 7) how a run begins, and 8) how outputs are evaluated. A trajectory records one run from the input task to the returned artifact. Because agents, operations, and initialization may be stochastic, repeated runs on the same task induce a distribution over trajectories rather than a single behavior. Our vocabulary turns structural design questions, such as when agents should communicate, gain or lose a capability, or carry information across runs, into testable choices. It also makes the evaluator a component of the system, since reported gains depend on how closely the proxy score matches true quality. That separation also splits the vague complaint that these systems lack taste into two failures with different solutions. Generative taste is the rate at which a system proposes novel trajectories before any score is observed, and evaluative taste is the gap between the proxy score and the quality it should match. We instantiate the vocabulary on recent autoresearch systems to illustrate that it covers designs that differ widely in structure.

Imprompt: A Language Framework for Prompt Programming

With the unprecedented success of Language Models (LMs), the science of Prompt Engineering has evolved the powerful idea of Prompt Programming, where prompts are treated as a programmable control surface for describing complex tasks and leveraging LM capabilities. However, existing prompt programming frameworks suffer from various complexities and inelegances, which make them hard to utilize in practice for effectively describing tasks. We propose Imprompt, a new language framework for the study and practice of prompt programming. We undertake a foundational investigation of prompt programming, and contend that prompt programs must contain only the task descriptions and must be decoupled from lower-level 'execution' details. We further develop this position by illustrating structured prompting as a combination of prompt programming and prompt program 'compilation'. We exemplify this view by formally defining two compilers for Imprompt programs. We then explore the idea of typing for prompt programs and draw a correspondence between type checking and constrained decoding. Finally, we implement our compilers and type checkers and evaluate them on a variety of case studies. We believe our work contributes programming-language foundations toward the emerging area of prompt programming.

Co-Harness: Co-Evolving Harnesses and Model Weights for LLM Agents

Post-training agents for automated AI research requires optimizing not only model parameters, but also the runtime harness that shapes how research trajectories are generated, evaluated, and learned from. Existing pipelines typically train models under a fixed harness, including prompts, tools, skills, middleware, and memory, while leaving the data-generating process outside the optimization objective. This creates a mismatch between model updates and the static scaffolding that determines trajectory quality. We introduce Co-Harness, a framework that jointly optimizes the agent harness and model parameters during post-training. Co-Harness alternates between harness optimization and model optimization. An LLM-based HarnessCritic analyzes failed trajectories, identifies harness-level failure modes, and proposes validated local updates. The model is then fine-tuned on high-quality trajectories generated by the improved harness, distilling effective scaffolding into model parameters. A 200+ hour autonomous case study further shows that Co-Harness can recover from system crashes, improve inference efficiency, and discover ensemble strategies without human intervention. These results suggest that joint harness and model optimization is an effective way to improve agents beyond fixed-harness post-training.

Beyond Sequential Interaction: Benchmarking Parallel Execution and Coordination for GUI Agents

Graphical user interface (GUI) agents are systems powered by large multimodal models (LMMs). They perceive screen state and execute user instructions through GUI actions such as clicking, typing, and scrolling on desktops and mobile devices. However, current agents scale poorly to long-horizon tasks: actions incur costly LMM inferences, and performance degrades as context grows. Humans divide such workloads among collaborators who complete sub-tasks in parallel. Yet parallel coordination among GUI agents has received little attention. To close this gap, we introduce ParaGUIBench, to our knowledge, the first benchmark dedicated to parallel execution and coordination of multiple GUI agents on separate desktop instances. It consists of three components: a multi-device Docker infrastructure with a shared file system; a dataset of 233 tasks spanning six task categories; and an evaluation system with efficiency metrics, including step reduction ratio and token cost. We further introduce ParaGUI, a planner-worker agent that decomposes GUI tasks and dispatches sub-tasks to concurrent workers on separate desktop instances. On ParaGUIBench, ParaGUI reaches a 46.4% success rate, outperforming the strongest serial baseline (Claude Sonnet 4.6) by 12.9 points while using roughly half the steps and less than half the tokens. These results show that parallel execution can improve both success rate and efficiency on decomposable, long-horizon GUI tasks, pointing to a direction worth further study.

LazyMem: Retrieve Broadly, Construct Selectively for Efficient Long-Term Agent Memory

Long-term memory lets LLM agents reuse past interactions, but raw dialogue histories are verbose and information-sparse. Retrieving broadly improves evidence coverage yet overwhelms downstream reasoning with noise; compressing at write time reduces noise but irreversibly discards details the future query may need. We introduce LazyMem, which sidesteps this dilemma by deferring all memory construction to query time. A lightweight 4B model processes the retrieved candidate pool in overlapping parallel windows, selectively retaining and compressing only query-relevant content. The model is trained through supervised fine-tuning followed by group-based reinforcement learning with a format-gated composite reward that combines a rule-based action signal measuring selection accuracy with an LLM-judged quality signal measuring source faithfulness and query utility. On the LongMemEval benchmark, LazyMem-4B achieves an LLM-judge accuracy of 0.85 with only 213 memory tokens, 68.7$\times$ fewer than retrieval-only, and generalizes to LoCoMo (0.68) without target-domain training, while reducing mean latency over the prior query-time baseline. The 32B variant reaches 0.93, surpassing oracle-context references on aggregation-heavy question types. The code associated with this work is publicly available at https://github.com/allacnobug/LazyMem.

HiLLTS: Zero-Shot Hierarchical LLM-Guided Traffic Signal Control for Sustainable Transportation

Urban traffic congestion significantly increases fuel consumption, greenhouse gas emissions, and commuter delays, resulting in substantial economic losses and environmental harm in modern cities. Traditional traffic signal control strategies such as fixed-time scheduling, actuated control, and reinforcement learning (RL)-based methods, offer different degrees of adaptability; however, RL-based methods can require extensive retraining, careful reward design, and substantial simulation data when transferred across networks or demand regimes. To address these challenges, we propose HiLLTS, an LLM-guided traffic signal control framework that employs a hierarchical three-layer architecture consisting of a central coordination agent, a district layer and multiple cluster-level intersection agents. Experimental results demonstrate consistent improvements in both congestion and environmental performance. Compared with the strongest non-LLM baseline in each scenario, HiLLTS reduces average waiting time by 36.73% under the low-congestion scenario and 14.71% under the high-congestion scenario, while reducing average CO2 emissions by 7.87% and 8.57%, respectively. Larger gains are observed against weaker baselines: under low congestion, HiLLTS achieves reductions of up to 18.00% in emissions and 62.07% in waiting time relative to Fixed-Time control; under high congestion, reductions of up to 28.89% in emissions and 40.36% in waiting time are observed relative to Max Pressure. The ablation study further validates the contribution of LLM-guided coordination over rule-based control

Risk Governance for Generative AI Mental Health Support: A Multi-Turn Safety Architecture

Large language models (LLMs) are increasingly used for emotional support despite lacking mechanisms to safely govern evolving mental health risk. Existing safety approaches primarily detect risk but rarely shape how models respond as conversational risk unfolds. We developed a model-agnostic safety governance architecture that combines contextual risk detection, reasoning-based verification, and protocol-guided response generation for multi-turn mental health interactions. Synthetic conversations grounded in real-world mental health narratives were used to evaluate the architecture's performance, tested with GPT-5-chat and Qwen3.5-27B, achieving high risk detection performance (specificity: 0.85 (95\%CI: 0.78;0.91), sensitivity: 0.92 (95\%CI: 0.88;0.95)) and increasing clinician-preferred escalation responses by 25.6--59.2pp while preserving rapport and connection. Performance remained stable across conversation length and generalized across both proprietary and open-source models. These findings demonstrate that clinically-grounded safety governance can extend beyond risk detection to improve how LLMs manage evolving mental health risk, providing a scalable framework for safer deployment across models.

Bayesian Repetition Penalty: A Principled Adjacent-Conditional Framework for Reversing Attention Collapse in Autoregressive Language Models

Attention collapse in autoregressive language models -- manifested as repetitive token loops where the model becomes trapped in self-reinforcing attractors -- is a persistent pathology that existing decoding-time heuristics fail to address at its root cause. We present a principled framework that penalises or compensates anomalous confidence arising from collapsed generation patterns, by comparing a token's observed frequency against its corpus prior through an adjacent-conditional probability construction. The resulting self-normalising penalty ratio $R=f(m,n,p)/f(np,n,p)$ requires no ad hoc standardisation and admits a closed-form logit offset with zero approximation error. The correction is isolated from the loss gradient and accumulated into a frozen output-layer bias via exponential moving average, enabling deployment as a repair mechanism for models that have already collapsed without requiring intrusive modifications to standard training pipelines. Experimental validation on a 1.5B-parameter model demonstrates that the frozen-bias mechanism can rescue a model already trapped in a collapsed attractor, reducing 2-gram repetition from 0.073 to near 0 while preserving generation quality.

PANOPTICON: A PII-Based Assemblage of Naturalistic Output Tokens for Investigating Privacy Leakage Within LLM Context Window

Large Language Models (LLMs) are capable of generalizing human language for the completion of never-before-seen tasks, leading to widespread deployment. While this automation provides clear utility, completing these tasks often requires the insertion of Personally Identifiable Information (PII), strings of information that uniquely identify some individual, raising privacy concerns. However, ethics has prevented the curation of a public, authentic dataset of PII. Without an appropriate dataset, it is difficult to quantify privacy risks. Thus, we introduce the PANOPTICON pipeline and dataset. The dataset, generated by Meta's Llama-3.1-8B-Instruct model, contains 67, 718 prompts, intended for the models context window, containing PII spans derived from 9,674 publicly available synthetic user profiles. We measure lexical diversity and S-BERT diversity of the created dataset to evaluate realism. Finally, we present a case study showcasing the utility of PANOPTICON data for understanding Prompt Inversion Attacks (PIAs). PANOPTICON thus emerges as the first benchmark dataset for studying PIAs over private corpora, providing a foundation for future LLM privacy research.

Test-Time Coverage: Test-Conditioned Data Curation for Deployment-Aware Learning

Deployed AI systems are often trained from broad candidate data pools, necessitating data curation towards the deployment test distribution. However, standard data curation methods score training-side criteria rather than directly optimizing deployment match. We introduce TTCov (Test-Time Coverage), a data-level test-conditioned curation method that uses test-side information before training instead of updating model weights at inference. TTCov decomposes deployment-conditioned curation into coverage and distribution. To represent coverage, it builds a task Atlas, a collection of LLM-based atomic propositions (APs) describing deployment-relevant concepts, seeded from open task knowledge and expanded with unmatched APs extracted from unlabeled deployment samples. To represent distribution, it instantiates the matched deployment APs with their frequencies, yielding a Knowledge Atlas (K-Atlas) that operationalizes the deployment distribution as a curation target. TTCov then selects a budgeted training set whose deployment APs distribution approximates this target. We apply TTCov towards autonomous driving (AD), keeping adaptation off the inference path while selecting data with greater deployment-relevant coverage, closer K-Atlas matching, and stronger downstream end-to-end driving performance than data-curation baselines, including seamless adaptability to novel domains via city-to-city expansion.

Similarity All The Way Up: Multilingual Generalization in LLMs Relies on Language-Level Similarity Structures

As Large Language Models (LLMs) grow more capable across diverse tasks, their (in)ability to generalize remains difficult to quantify and poorly understood beyond limited domains. In particular, LLMs are known to struggle generalizing multilingually, to languages outside of English, and that are poorly attested in their training data. To understand why this may be, and what enables some models to perform better than others, we turn to a long history of work across the cognitive sciences, arguing that successful generalization derives from appropriate representations in similarity space. We look at how well LLMs' representations capture the hierarchical similarity structure between distinct languages. Strikingly, we show LLMs' latent representations largely recover the hierarchical structure of the Indo-European language family tree -- grouping languages that are members of the same subfamily closely together in representation space. Furthermore, we show that the degree to which models reflect the similarity structure of languages correlates with their performance on XNLI, a multilingual natural language inference benchmark. This extends classic work on similarity-driven generalization at scale, showing how models that represent similar languages similarly generalize better from one language to another.

RoleMix: Unifying Sequential and Non-Sequential Features via Semantic Tokenization for Post-Click Conversion Rate Prediction

Post-click conversion rate (PCVR) prediction is central to industrial recommendation, but remains challenged by the structural mismatch between sparse, unordered multi-field features and long, domain-specific behavior histories. Existing models often process these signals through separate pathways and fuse them late, weakening semantic roles and limiting cross-signal refinement. We propose RoleMix, a unified interaction architecture that represents sequential and non-sequential evidence through a shared, role-preserving token interface. Non-sequential fields are converted into explicit semantic tokens that preserve user, item, pairwise, dense, contextual, and cross-feature roles, while long behavior domains are compressed into item- and context-aware sequence-query tokens through two-stage hierarchical window attention. The resulting global, semantic, and sequence-query tokens are jointly refined by stacked UniMixing-Lite blocks for PCVR prediction. On the large-scale KDD Cup 2026 Tencent UniRec Challenge, RoleMix achieves 83.648% online AUC, outperforming the official industrial baseline by 1.953%. Ablation studies show that semantic tokenization yields the largest isolated gain, highlighting a key principle for large-scale PCVR modeling: preserving field semantics at the token-interface level is as important as scaling the interaction backbone.

MPR-CiteG: Enhancing RAG with Multi-Portfolio Retrieval and Citation-Grounded Generation

This paper presents the MPR-CiteG framework, which achieved second place in the ScienceON AI Challenge by addressing two fundamental challenges in generative AI: inefficient retrieval and the absence of source verification. We propose a dual-component system, termed MPR-CiteG, in which the Multi-Portfolio Retriever (MPR) efficiently retrieves diverse and relevant information, while the Citation-Grounded Generation (CiteG) module ensures that every generated output remains factually consistent and explicitly attributed to its source. MPR-CiteG represents a significant step toward building more trustworthy and accurate LLMs that are not only capable of generating information but also of grounding their responses in reliable evidence, thereby mitigating common issues like model hallucination. Extensive experiments on the challenge dataset validate the effectiveness and reliability of our approach. Our code is available at https://github.com/2noweyh/MPR-citeG.

SEGRA: Structured Experience-Guided Graph Reasoning Agent for Gremlin Based Question Answering

Enterprise IT support knowledge graphs capture rich relationships among cases, users, devices, symptoms, taxonomic categories, root causes, and historical resolutions. Yet querying them in Gremlin requires knowledge of graph schemas, traversal semantics, edge directionality, and property-graph-specific constraints, making them difficult for non-expert operators to use. We introduce SEGRA, an experience-guided agent for enterprise text-to-Gremlin question answering. SEGRA integrates intent routing, schema- and taxonomy-grounded query generation, multi-shot decomposition, execution-aware verification, and a curriculum-bootstrapped skill library that reuses verified query patterns. On an enterprise IT support benchmark, SEGRA achieves a $7.0\times$ higher mean judge score than backbone-only chain-of-thought prompting. Its skill library further reduces LLM calls by $20\%$ and dollar cost by $18\%$ relative to SEGRA without skills, while preserving answer quality. These results show that schema-grounded agent design and reusable execution experience improve both accuracy and efficiency for enterprise graph QA.

Spatial Reasoning in LLM Game Agents: Impact of Causal Context and Multi-Step Planning

LLM-based game agents often perform poorly on more complex tasks. This work examines whether these failures are linked to limited spatial reasoning and evaluates whether causal prompt augmentation and multi-step planning can improve win-rates while managing response latency. Using the open-source Qwen3 model family, we conduct experiments across varying model scales, reasoning modes, and planning horizons. We further introduce a focused GVGAI benchmark consisting of three custom games with five difficulty levels to isolate spatial navigation. The evaluation follows two paradigms: an initial ``positioning experiment'' to test an agent's ability to find its exact coordinates, and a study of game-play success. Our results show that while larger models with an enabled thinking mode identify their positions more accurately, overall performance in coordinate matching remains limited for smaller models. Win rates decrease as game levels and layout complexity increase, validating the benchmark's difficulty scaling. Integrating causal context into the prompts tends to improve the agents' success rates, particularly for bigger models. While enabling thinking mode and longer planning horizons significantly improve performance, multi-step planning further reduces mean per-step response times, offering a practical trade-off between reasoning depth and execution speed.

Commitment To Cooperation With Self-Negotiated Contracts

As AI agents operate with increasing autonomy in a multi-agent world, they will need to learn to cooperate with other agents and with humans to generate mutual benefits. However, cooperation is a challenge because the costs of cooperation are often incurred early on, but the benefits are only realized later, creating an incentive to defect. How can AI agents cooperate with commitment? Here, we draw on inspiration from legal institutions and contracting that human societies have used to solve principal-agent problems of this kind. Contracts provide observable representations of agreements that enable credible commitments through the enforcement of terms. We study the role of contract-based cooperation using LLM-based agents in \CT, a spatial-temporal game that combines bargaining with navigation towards a goal. We study a suite of contract representations that range from formal contracts that compile to code to natural contracts that require reinterpretation. We evaluate agents with a range of LLM backbones using different sizes and providers. We find that self-negotiated contracts can improve cooperative outcomes beyond what is possible with regular trading.

Disentangling Multi-View Scanning in Mamba for Network Traffic Anomaly Detection

Network Traffic Anomaly Detection (NTAD) is a critical task in cybersecurity, yet timely and accurate anomaly detection remains challenging. Mamba has emerged as a particularly promising backbone for NTAD due to its linear-time complexity for long-sequence modeling. It further incorporates a dedicated multi-view scanning mechanism to enhance detection precision through complementary contextual cues. However, we identify a previously overlooked structural deficiency in multi-view Mamba scanning for NTAD: redundancy accumulation. Specifically, distinct scanning branches capture substantial view-invariant information, which is repeatedly amplified during multi-view fusion; conversely, view-specific information is diluted or even suppressed, leading to representation homogenization and multi-view degradation. To address this problem, we propose DisenMamba, a novel disentangled multi-view Mamba framework. DisenMamba reformulates multi-view scanning as a two-stage disentangle-then-fuse process that explicitly separates view-invariant and view-specific components prior to fusion. This design prevents the invariant information accumulation while preserving complementary multi-view cues, yielding more discriminative representations for subtle traffic anomalies. Extensive experiments demonstrate the effectiveness of DisenMamba, establishing a new disentangled multi-view Mamba paradigm. Code is available at https://github.com/ikun0124/DisenMamba.

Coordinated Networking for On-Device Agent-Augmented Real-Time Communication

AI agents are enabling a new paradigm of agent-augmented real-time communication (RTC), where humans focus on high-level collaboration, while agents autonomously retrieve, analyze, and generate information in real time to support their interactions. These apps enable new experiences across various domains: for example, when corporate employees co-author a legal document, their agents can discuss and draft on their behalf, sparing them the burden of manually reviewing each other's work. As existing cloud-based agents suffer from privacy risks and unscalable server costs, on-device agent-augmented RTC offers a promising alternative. However, this on-device paradigm introduces a new networking challenge: contention between concurrent traffic flows generated by humans (for live video streaming) and agents (for sending context files for analysis). We design HFS, a framework to ensure both high live video quality and low agent response latency in agent-augmented RTC apps. We achieve the goal through an app-guided multi-flow transport approach, where a unified app-layer orchestrator jointly controls the sending rates of live video and agent context flows based on their heterogeneous app requirements. Our prototype built atop WebRTC and llama.cpp demonstrates that HAFS outperforms baselines, achieving 1.5x higher video quality while reducing agent response time by 31%.

What Can Be Enforced? A Theory of Certified Runtime Safety for Tool-Using Agents

Runtime guardrails act before irreversible tool calls, but their guarantees depend on what policy state is representable, what a judge observes, and whether intervention changes future behavior. We separate three questions. First, relative to fixed oracle predicates, a deterministic gate enforces exactly the nonempty safety policies whose good prefixes its register model recognizes; policy nontriviality is undecidable with two decrementable counters but in PSPACE for a separable monotone fragment. Second, under a fixed exogenous law, Neyman-Pearson gives the exact false-block/miss frontier and conformal calibration gives a finite-sample marginal certificate, possibly via block-all. Third, once blocking changes future proposals, static scores and ungated trajectories need not identify the closed-loop frontier; a specified finite controlled model instead yields an occupancy program. Bounded representation attacks add a robustness margin, so benign calibration alone does not transfer. Experiments target these distinctions through static diagnostics, controlled-model enumeration, representation rewrites, and paired closed-loop reruns.

Physical AI Governance: From Theory to Practice Across Life Cycle

With the emergence of Physical AI, artificial intelligence is extending beyond screen-based applications to embodied systems that perceive, interact with, and act in the physical world. Unlike traditional AI, Physical AI operates under real-time safety constraints, continuously interacts with dynamic environments, and coexists with humans, introducing governance challenges that existing AI governance frameworks do not explicitly address. This paper presents a comprehensive survey of Physical AI governance from both scientific and operational perspectives. We synthesize existing governance principles and organize them into a unified governance framework tailored to physical AI systems. Building on this foundation, we propose a five-stage Physical AI lifecycle comprising research, design, data, model development, and deployment, and demonstrate how governance can be operationalized across each stage through concrete implementation practices. By connecting governance principles with engineering workflows, this survey provides a structured reference for researchers, developers, and policymakers to build Physical AI systems that are safe, trustworthy, and aligned with societal values.

How Well Can AI Generate Backlogs from App Mockups?

Creating sprint backlogs requires considerable effort, as items such as epics, user stories, and tasks can be missed or inconsistently specified. We propose a multimodal approach to support backlog generation from visual app mockups, an artifact available at early project stages. We evaluate three prompting strategies on GPT-4o: a zero-shot baseline, Compositional Chain-of-Thought (CCoT) for vision-language reasoning, and a persona-driven prompt. We study seven app development projects across two countries and interview developers about the results. Overall, we observed that the baseline prompt favours recall over precision, whereas CCoT is more balanced, achieving average F1 scores of 52-66% for epics and user stories. Tasks were more challenging to generate accurately. Precision gains were most consistent when adding architectural context, particularly for backend tasks (precision gains up to 35%). Interviews with developers revealed that up to 26% of false positives were still considered useful, reflecting the creative and open-ended nature of backlog creation. To capture this, we propose a new measure called Revised Recall, which complements ground-truth evaluation with developer assessments. Our findings suggest that hybrid prompting with architectural context can assist backlog generation from early mockups, though results vary by item type and developer oversight remains necessary.

Agent Team Work Zone: An Automated, Persistent Workspace for Long-Lived Coding Agent Teams

Large Language Model (LLM) agents have significantly improved coding and programming workflows. Claude Code, in particular, is one of the most powerful LLM coding agents and is capable of conducting complex coding tasks. However, several drawbacks can undermine long-term agentic workflows. (1) Irrecoverable agent teams: The Agent Teams feature is powerful, but the working state accumulated by each teammate is lost and cannot be resumed once the process stops, for example, when a terminal is closed. (2) Compaction erodes working detail: Compaction condenses the conversation into a summary, causing an agent's working details to become vague. (3) Agentic "technical debt": Over time, a user's decisions and the agents' operations become trapped in compacted old chats, making the project increasingly difficult to maintain and review. (4) Heavy prompt writing: Assigning or handing off tasks requires users to repeatedly write long prompts to achieve the expected agentic performance. We propose ATWZ (Agent Team Work Zone), a filesystem-based operations layer built around Claude Code's native Agent Teams that addresses these problems. Its central design principle is to treat each agent and teammate as a human employee and preserve their important working state in files stored in a dedicated directory called a "workstation," together with the skills, hooks, and scripts that use and maintain these files. With ATWZ, an agent team can periodically back up its working state, allowing an agent's knowledge to be recovered after compaction. After a process ends, the team can be restored with a single command. These features also substantially mitigate the agentic "technical debt" described above. Moreover, within ATWZ, agent "employees" can send documents to one another, greatly reducing the effort required to write prompts.

SAGE: Safety-First Defense-in-Depth Guardrails for Verified Lifecycle Control of High-Impact Generative AI

High-impact generative AI makes catastrophic misuse a lifecycle-control problem, not merely a prompt-filtering problem. SAGE is a safety-first, authorization-separated architecture in which credible catastrophic-enablement risk constrains admissibility before utility, latency, or commercial objectives are considered. It combines signed release manifests, diverse detectors, robust risk envelopes, least-risk defaults, output checking, three-valued monitoring, protected audit chains, containment, and rollback. Formal results establish safety priority, conservative detector bounds, monotone release gating, tamper-evident records, and an authorization cut; two PRISM abstractions verify authorization separation and lifecycle invariants under explicit assumptions. A frozen, vendor-symmetric study sent 84 cases to each of four GPT, four Claude, and two Gemini snapshots: 840 calls yielded 794 target responses, 46 provider errors, and 449 successful judgments covering 375 responses. Eight snapshots had complete judged domain coverage. Harmful-compliance estimates were low; variation arose mainly from benign utility and safe redirection. Seven multiplicity-adjusted contrasts involving Claude, Gemini, or GPT-5 snapshots and the GPT-5 mini and GPT-5 nano snapshots were supported, while no tested contrast between the Claude or Gemini snapshots and GPT-5 or GPT-5.5 survived correction. The observed harmful-compliance range is a conservative, protocol-bound view from one generation per prompt with no tools, retrieval, history, or human adjudication; it is not an upper bound on operational assistance. A preregistered extension specifies how to test a wider best-worst gap using a locked split, repeated sampling, multi-turn and sandboxed-tool conditions, and domain-expert scoring.

Design Theater: A Benchmark for Generative UI

Generative UI tools promise to democratize UI design by turning natural language descriptions into complete interfaces. Alongside the interface, these tools generate user-facing design rationales that explain their layout, accessibility, and design choices. However, it remains unclear whether these stated rationales are actually reflected in the interfaces they produce. We call this disconnect ``Design Theater'': plausible and confident design rationales that have little relationship to the actual implementation. To study this phenomenon, we introduce a benchmark and three metrics for measuring Design Theater. The benchmark includes 24 UI generation tasks spanning structural, styling, and functional design requirements. Using this benchmark, we evaluate 120 interfaces created by five generative UI tools. On average, over 25\% of user-facing design rationales are not implemented in the generated interface, and the implementation failure increases to 34\% for functional requirements. Tools recognize roughly half of the UX principles embedded in prompts (mean = 0.54), with four of five tools implementing 6\% or fewer functional principles. We also measure interface similarity across tools and find convergence in visual appearance and layout organization, with greater variation in color choices. Overall, we contribute: 1) the concept of Design Theater; 2) a benchmark with metrics for assessing whether the stated reasoning of generative UI tools is reflected in their implementations; 3) and findings from a systematic evaluation of these tools. We discuss what these findings mean for the design and evaluation of generative UI tools.

Let AI Agents Translate Networks, Not Reason About Them

A formal model enables verifying reachability, localizing an outage, or anticipating the blast radius of a change. Yet, virtually no production network has one, since writing a model by hand demands rare expertise and is hard to keep current as the network changes frequently. At its core, network modeling is a typographical exercise: it translates network artifacts (e.g., configurations, topology, and routing state) into rules in formal logic. Translation of this kind is what large language models (LLMs) nowadays do well. Unlike free-form AI reasoning, such translation can be formally verified. Once modeling is no longer the bottleneck, trusting AI to reason over large, complex networks no longer makes sense. Our position therefore cuts against the prevailing race to put autonomous AI agents in charge end-to-end. We instead confine AI to translation and rely on a solver for reliable long-horizon reasoning, building a reusable formal model of general network behavior that can then be specialized to specific tasks, e.g., root-cause analysis (RCA). We build TypoNet that constructs and validates a symbolic model of an emulated production-scale WAN from the network's own artifacts. Our preliminary evaluation shows TypoNet helps in two ways. On its own, TypoNet answers operational questions (e.g., reachability verification and change-impact analysis) faster, more cheaply, and more reliably than an LLM. As a tool for an AI agent, TypoNet boosts fault localization at lower cost. The result makes the case for AI that builds verifiable network models and relies on a solver for reliable long-horizon reasoning.

Share No More Than the Request Requires: Federated Disclosure for Perspective-Aware AI

Modern AI systems bring societal risks such as mass surveillance, extreme concentrations of power, and loss of user autonomy---calling into question a model where third-parties collect and control massive amounts of user data. Users require a sovereign system to securely own, govern, and disclose their context while remaining compliant across regulated domains with strict provenance, interpretability, and policy adherence. Perspective-aware AI approaches this by transforming a user's aggregated personal data into a structured identity model called a \emph{Chronicle}: a temporal knowledge graph that represents and grows with the user. Chronicles support the secure disclosure of context across federated networks. A Chronicle holder may expose a queryable, authorized view that a third-party agent may consult without centralizing anyone's data. This paper explores the problem of minimum-necessary disclosure across domain boundaries: when a requester's agent queries a Chronicle, how can the system constrain its response to release only what the requester's relationship, stated purpose, and specific task require? We propose \textbf{Provenance Preserving Chronicles} (PPC), a federated protocol that compiles each holder's Chronicle into a compact \emph{authorized evidence subgraph} governed by one rule: \emph{share no more than the request requires}. Holders keep local sovereignty; an access controller projects relationship-aware views over domain-expert ontologies; and a two-phase flow returns provenance-linked text first, releasing raw artifacts only after explicit holder approval. We frame the problem, map gaps in blockchain, P2P, and holder-sovereign designs, define the core constructs, and sketch the protocol with an explicit threat model.

ConsistencyGate: Preventing Memory Contamination in LLM Agents via Self-Consistency Admission Control

LLM agents that operate over many turns accumulate facts in an external memory store and reuse them as premises for downstream reasoning. A hallucinated fact written at one step therefore persists as a false premise for every subsequent step, a failure mode we call memory contamination. Existing memory management addresses retrieval and capacity but not write-time correctness; this admission problem cannot be solved by utility- or recency-based criteria, and uncontrolled contamination compounds across long trajectories. We propose ConsistencyGate, a write-time admission gate that, before committing a candidate fact m extracted from context c, queries the LLM K times for a soft support score and admits m only when the average exceeds a threshold. The mechanism is model-agnostic, requires no fine-tuning, and reduces to a single forward pass in a log-probability variant for latency-sensitive deployments. To measure the effect on natural data, we construct two real-conversation benchmarks (LoCoMo-Contam and MSC-Contam) by planting controlled single-detail corruptions in long-term conversations from LoCoMo and MSC, and complement them with a structured synthetic corpus (MemContam) that isolates a near-oracle upper bound. Across four LLM backbones, ConsistencyGate reduces contamination on every benchmark relative to a write-everything baseline, with the cost concentrated on facts that are stated only implicitly in the source context. We release all three benchmarks together with the gate implementation.

Reason Popper-ly: Patching In-Context Reasoning with Inductive Logic Programming

Chain-of-thought (CoT) prompting enables large language models (LLMs) to tackle multi-step reasoning tasks, yet the generated intermediate steps are not guaranteed to be logically sound. We present Reason Popper-ly, a neurosymbolic framework that uses inductive logic programming (ILP) to learn relation composition rules from reasoning traces and deploys them as an online verifier for step-level correction. Given an LLM-generated trace, the method checks each inferred step against the learned rule table, diagnoses the violation type, rewrites incorrect steps with symbolically derived repairs, and regenerates the remaining suffix so that the model can produce its final answer conditioned on a verified trace. We evaluate on CLUTRR, a multi-hop kinship reasoning benchmark, using five language models over reasoning chains of 2 to 10 hops. Across all models, Reason Popper-ly consistently improves terminal accuracy over standard CoT, with gains of up to 48 percentage points for small models and 15 points for frontier models on the longest chains. Compared with a fully exogenous symbolic pipeline, our method performs better on harder instances by preserving the model's successful grounding while correcting only verifiable reasoning failures. In addition, step-level ILP verification yields a fine-grained error taxonomy that provides diagnostic insight beyond final-answer accuracy.

LiteResearcher: A Scalable Agentic RL Training Framework for Deep Research Agent

Reinforcement Learning (RL) has emerged as a powerful training paradigm for LLM-based agents. However, scaling agentic RL for deep research remains constrained by two coupled challenges: hand-crafted synthetic data fails to elicit genuine real-world search capabilities, and real-world search dependency during RL training introduces instability and prohibitive cost, which limits the scalability of Agentic RL. LiteResearcher is a training framework that makes Agentic RL scalable: by constructing a lite virtual world that mirrors real-world search dynamics, we enable a continuously improving training recipe that empowers a tiny search agent to outperform large-scale open-source and commercial models (e.g., Tongyi DeepResearch and Claude-4.5 Sonnet). Specifically, on common benchmarks such as GAIA and Xbench, our LiteResearcher-4B achieves open-source state-of-the-art results of 71.3% and 78.0% respectively, demonstrating that scalable RL training is a key enabler for Deep Research Agents.

What Does Chain-of-Thought Contribute at Probe Time? Evidence for Local Co-Occurrence Activation

Chain-of-thought (CoT) prompting enhances large language model performance, yet what drives these gains remains unclear. We study this question from a probe-time perspective: holding CoT rationales fixed, we test which textual properties matter for the final prediction. Across multiple datasets and model configurations, we find that randomizing the order of rationale sentences has little effect on accuracy, suggesting that the global order of reasoning steps is not the main source of the probe-time benefit. Moreover, even when the words in a rationale are randomly reordered, performance remains well above the no-rationale baseline, indicating that the rationale's words remain useful even without their original order. Restoring only short-range word order further improves performance and brings it substantially closer to full CoT. In most settings, much of this local-order gain is already obtained with three-word windows. Control experiments rule out explicit answer copying, simple lexical cues, generic topical context, and general robustness to shuffling as the main explanations. Mechanistic analyses further show that short-window gains are largely formed in early-to-middle model layers, with answer-relevant evidence concentrated in local text spans. Together, these findings support a local co-occurrence activation (LCA) interpretation: the probe-time benefit of fixed rationales arises mainly from the words they contain and short-range word co-occurrences.

On the Detection of Commutative Factors in Factor Graphs: Necessary and Sufficient Conditions

Exploiting the indistinguishability of objects in a probabilistic graphical model such as a factor graph is key to lifted probabilistic inference algorithms and allows for tractable probabilistic inference problems with respect to domain sizes. A central building block for the exploitation of indistinguishable objects in factor graphs is the identification of commutative factors, i.e., factors whose output values are invariant under permutations of input values assigned to a subset of their arguments. In this paper, we revisit the theoretical foundations underlying the state-of-the-art algorithm to detect commutative factors. Specifically, we show that in its current form, the state-of-the-art algorithm relies on a central theorem that is mistakenly regarded as a sufficient condition to identify commutative factors, while it actually only implies necessary condition. Consequently, the state of the art might, as we show in this paper, deliver incorrect results. To fix the flaws currently present in the state of the art, we prove a slightly modified version of the aforementioned theorem, which serves as a necessary condition to identify commutative factors. Moreover, we present a corrected version of the state-of-the-art algorithm, which keeps its efficiency while ensuring correctness and introduce a complementary algorithm with tighter worst-case bounds.

WIRE: Profiling Witnessed Within-Policy Instruction Collisions in LLM Agents

LLM agents are governed by long-lived prompt policies, where individually reasonable stand- ing rules can jointly govern the same pre- generation state. Existing instruction-following evaluations usually ask whether a model satis- fies explicit constraints, but they do not show how a model resolves pressure among rules inside one standing policy. We introduce WIRE, a witnessed resolu- tion profiler for prompt policies. WIRE ex- tracts source-grounded rules, encodes them as PYRULE clauses, uses satisfiability checks only to nominate same-surface hard-collision can- didates, realizes those candidates as concrete co-governance witnesses, and executes subject models to produce a four-cell resolution profile: satisfy both rules, only the earlier rule, only the later rule, or neither. Across six public prompt policies, WIRE ex- tracts 276 source rules and 560 clauses, clas- sifies 30,944 within-policy clause-pair com- parisons, retains 170 encoded hard-collision source-rule pairs, and realizes 1,402 concrete witnesses. In policy-only evaluation, these wit- nesses yield 13,335 jointly governed, judgeable trials; only 35.4% satisfy both governed rules. The resulting profiles reveal policy-specific, model-specific, and tool-interface-specific res- olution patterns. WIRE is not a proof of natural-language contra- diction, a deployment-frequency estimator, or a root-cause diagnosis. It is a measurement tool that returns reproducible witnesses and aggre- gate profiles for inspection, regression testing, and repair.

Universal Quantum Transformer

Classical continuous-space neural networks fundamentally struggle to lock into exact formal rules, whether mathematical, such as modular arithmetic and non-Abelian group algebra, or linguistic, such as systematic compositional generalization. To approximate these discrete logical rules, they often rely on massive parameter scaling, resulting in stochastic instability even after delayed generalization phenomena known as grokking. Here, we introduce the Universal Quantum Transformer (UQT), a novel, quantum-native computing architecture that uses the physical properties of multi-qubit systems as a universal inductive bias for exact algebraic and compositional reasoning. Rather than translating classical neural mechanisms, our framework relies entirely on parameterized geometric phase embedding and $SU(2)$ wave-interference. We demonstrate that an identical quantum attention circuit, operating on a highly compact 5 or 6 qubit substrate with only 551 to 1,650 trainable parameters, exactly learns three highly distinct formal classes: cyclic modular arithmetic ($\mathbb{Z}_{11}$), non-Abelian algebra (the $S_4$ permutation group), and systematic linguistic compositionality (the SCAN language). While standard classical models, including multi-layer perceptrons (MLPs) and Transformers, exhibit stochastic instability at convergence, the UQT achieves mathematically exact, deterministic generalization. We define this stricter regime as crystallization: a step beyond the well-known phenomenon of grokking. Finally, we deploy the UQT on noisy intermediate-scale quantum (NISQ) hardware, achieving 97.5% accuracy on IBM Quantum computers. These results demonstrate that the UQT provides a structurally suited inductive bias for exact formal reasoning that standard classical continuous-space architectures do not natively provide.

Tree-Based Formalization of Multi-Agent Complementarity in Human-AI Interactions

Complementarity is the case in which a human--AI interaction (HAI) outperforms the best prediction benchmark available among its members. Although this idea is central in HAI research, formal work on complementarity remains limited. Existing frameworks do not model how agents' predictions compose into workflow-sensitive multi-agent protocols. We close this gap by introducing a tree-based formalization of complementarity in multi-agent HAI. An HAI protocol is represented by an ordered agent-role configuration together with a rooted planar binary tree whose leaves are decorated by prediction vectors. A local binary composition rule is evaluated recursively along the tree, yielding a tree-relative complementarity functional relative to a pointwise-min benchmark. We prove four results. First, selector-based HAIs, including reliance, cannot achieve complementarity regardless of task, loss, or prediction quality. Second, in regression under squared loss, complementarity is equivalent to Euclidean distance minimization from the ground-truth vector; for $N=2$, the optimal linear-pooling weight has a closed form and a residual-correction interpretation. Third, under linear local composition, every protocol tree defines a barycentric coordinate chart on the simplex of leaf weights; Tamari-cover reparameterizations of protocol trees preserve complementarity, and for all $N$, any two Tamari paths with the same initial and terminal trees preserve the protocol output and complementarity. Fourth, in binary classification, no internal local composition can achieve complementarity under endpoint-monotone losses, including standard Bregman and many finite Bernoulli $f$-divergence losses; an analogous obstruction holds for multiclass aggregation under cross-entropy. In summary, our framework shows that complementarity is attainable in multi-agent regression, but obstructed in classification.

Some hypotheses on how chatbots work in problem-solving-driven conversations. Large Language Models as confirmation of the Innovation Illusion

We discuss the nature of chatbots as conversation partners in problem-solving. What can chatbots do and what can't they do? We develop hypotheses on how this can this be explained. Our argument draws on insights from Aggregation Dynamics, Cognitive Linguistics, Neuropsychology and Psychology. We establish that chatbots are multifaceted and composite systems. Our argument focuses on basic chatbots in the hope of thereby making statements about the core functionality of more advanced chatbots. Basic chatbots are assumed to consist of a Large Language Model (LLM) with a simple interface. The main results of our research are: a description of human imagination, understanding and thinking based on so-called metaphorical problem propagations; the hypothesis that the texts in the text dataset used for training LLMs have specific characteristics and that these texts only partially imitate human thinking and understanding; the hypothesis that the LLM training process encodes artificial metaphorical problem propagations into an LLM from these text datasets. Our conclusions are that a basic chatbot cannot be a thinking partner capable of matching the cognitive flexibility of humans, and that further development of the Large Language Model will not lead to this either. But chatbots exist, they are being used on a massive scale, by both individuals and organisations. It is therefore socially and politically important to understand them. Our article aims to contribute to the discussion on the functioning, benefits and drawbacks of chatbots. Cognitive Linguistics shows how the use of metaphor is an expression of our thinking. Aggregation Dynamics, is an attempt at a comprehensive systems theory. We believe that the concept of metaphorical problem propagation could provide an interesting addition for both. Chatbots a solution? For what?

Artificial Intelligence for Mathematical Reasoning: An Integrated Survey of Language Models, Neuro-symbolic Systems, and Verified Discovery

Mathematical reasoning has long served as a stringent test of machine intelligence; over the past decade, it has moved from a niche problem within NLP to one of the most consequential AI frontiers. This survey provides a unified account of the field's evolution, from early rule-based math word problem (MWP) solvers and template-driven geometry systems, through neural expression generation and LLM prompting, to contemporary reasoning models, multi-agent systems, neuro-symbolic theorem provers, and verified discovery workflows. We organize the landscape along four axes: (i) informal reasoning over text and diagrams, spanning MWP solving, multimodal geometry, and VLMs; (ii) formal reasoning in proof assistants, including autoformalization, tactic prediction, compiler-guided repair, and proof search; (iii) mathematical discovery, where systems propose constructions, improve bounds, or assist attacks on open problems; and (iv) the inference and training-time techniques, including CoT prompting, tool use, process reward models, and RLVR, that increasingly connect generation with verification. We catalog major benchmarks across grade-school arithmetic, competition mathematics, geometry, formal proving, multimodal and multilingual reasoning, and expert evaluation, and we examine benchmark saturation, contamination, reporting mismatches, and the distinction between pass@1, majority voting, and verifier-assisted pass@$k$. We critically assess failure modes: brittleness under perturbation, reward hacking, multimodal grounding failures, fragile formalization, and the energy cost of reasoning-scale inference. Drawing on recent perspectives from working mathematicians, we identify future directions centered on verified-discovery workflows, reasoning efficiency, and infrastructure to make AI-assisted formalization broadly usable. Companion materials: https://github.com/Starscream-11813/awesome-AI4Math.

ProvenanceGuard: Source-Aware Factuality Verification for MCP-Based LLM Agents

Tool-using LLM agents increasingly use the Model Context Protocol (MCP) to answer from heterogeneous evidence sources, including search, APIs, databases, clinical records, and formulary tools. Standard factuality metrics usually test whether an answer is supported by pooled evidence, missing a provenance-sensitive failure mode: a claim may be supported somewhere while being attributed to the wrong source. We call this cross-source conflation. We introduce ProvenanceGuard, a source-aware verifier for MCP-grounded answers. It consumes captured MCP traces with stable tool IDs, source IDs, and raw outputs; decomposes answers into atomic claims; routes claims to source-specific evidence; checks support with NLI and a token-alignment proxy; compares stated attribution with the routed source; and returns per-claim verdicts plus an answer-level allow/block decision. Blocked answers can be repaired with retrieval-augmented answer revision and re-verified. We evaluate on 281 medical-domain MCP-agent traces. A 266-trace adjudicated subset yields 2,325 LLM-assisted claim labels split by trace; 361 held-out labels are human-verified. On the 40-trace held-out split, ProvenanceGuard achieves block F1 0.802 and source accuracy 0.858 over 260 source-eligible claims, outperforming source-blind baselines that do not emit claim-to-source IDs. On a harder multi-source benchmark it reaches block F1 0.846, while source-plus-relation accuracy drops to 0.229, showing that exact source ownership remains difficult with semantically close sources. Repair-and-reverify resolves all blocked answers in the full trace set, often via conservative fallback. In 50 controlled clinical conflation probes, ProvenanceGuard detects all injected attribution swaps with no retained wrong attribution. These results show that source attribution is an independent axis for factuality verification in MCP-based agents.

ATOD: Annealed Turn-aware On-policy Distillation for Multi-turn Autonomous Agents

Training small language-model agents for long-horizon interactive tasks requires both fast imitation and reward-driven improvement. On-policy distillation (OPD) provides dense teacher guidance and typically improves rapidly in the early stage, but its gains saturate once the student approaches the teacher, limiting the final performance ceiling. Reinforcement learning (RL) directly optimizes environment rewards and encourages exploratory improvement toward a higher reward-defined ceiling, but sparse and delayed feedback makes early-stage learning much less efficient than OPD. In this paper, we propose ATOD (Annealed Turn-aware On-policy Distillation), a hybrid online distillation algorithm that explicitly exploits this complementarity. (1) ATOD uses an annealed OPD-RL schedule: OPD dominates early training to approach teacher-level behavior, while RL is gradually strengthened to drive reward-based exploration. (2) ATOD introduces Turn-level Disagreement-Uncertainty Reweighting (T-DUR), which softly amplifies high-utility turns and improves dense supervision in long trajectories. Experiments on ALFWorld, WebShop, and Search-QA show that ATOD consistently outperforms competing post-training baselines: across the three student sizes, ATOD improves average success rate by 3.03 points over OPD and 23.62 points over GRPO, while surpassing the corresponding teacher models by 2.16 points.

EvalSafetyGap: A Hybrid Survey and Conceptual Framework for LLM Evaluation-Safety Failures

This paper presents a systematic survey and conceptual synthesis of the shared measurement problem underlying large language model (LLM) evaluation and AI safety: benchmark scores, reward signals, and safety metrics can improve while the capabilities and alignment properties they are meant to represent remain uncertain. Synthesizing 373 primary studies published between 2018 and 2026, the survey organizes evidence on benchmark validity, contamination, dynamic evaluation, LLM-as-a-judge protocols, adversarial safety testing, reward and proxy optimization, mechanistic interpretability, and AI governance into an eight-stream evidence taxonomy. Building on this synthesis, we introduce EvalSafetyGap, a conceptual framework that unifies benchmark-validity and alignment-failure research as a shared proxy-target divergence problem under optimization pressure, formalized through a Goodhart-inspired Instability Decomposition and an Alignment Trilemma. An exploratory ten-model public-evidence audit illustrates the framework by showing why capability, behavioral robustness, and governance disclosure should be reported as separate evidence layers rather than collapsed into a single safety score. The survey closes with a research agenda for dynamic and contamination-resistant benchmarks, pre-specified multi-attempt threat models, version-locked evaluation, transparent source reporting, and validated mechanistic safety indicators, offering researchers, model developers, and AI auditors a shared vocabulary for measurement-aware LLM safety evaluation.

Agents Don't Just Agree, They Remember: Benchmarking Persistent Sycophancy in Stateful Personal Agents

Stateful personal agents increasingly maintain long-term user profiles, episodic memories, and reusable skills. This persistence turns conversational sycophancy into a state-writing failure: accepted user-centric claims can be committed as lasting preferences, background facts, or workflows and later reused after the original conversation is gone. We call this persistent sycophancy and introduce the Personal Agent Sycophancy Benchmark (PASB), a 1,600-task benchmark that traces whether a conversational claim is accepted, written into durable agent state, and reused in a later neutral query. Unlike prior benchmarks that provide pre-written memories, PASB evaluates real agents (Hermes-Agent and OpenClaw) that decide what to store. It isolates the write process by combining four scenario framings with four temporal delivery patterns and separating a five-turn persist stage from a cleared three-turn query stage, ensuring downstream effects arise only from durable state. Across twelve models, the commit boundary is the key inflection point: downstream failure increases from 45.0% in session-only episodes to 71.9% after commitment, a consistent increase of 27.0 percentage points. Committed claims exhibit three write-time patterns: status promotion, attribution removal, and scope broadening. These patterns become stronger under memory-like or procedural framing, repeated reinforcement, and even across domain boundaries. These results show that agent sycophancy is fundamentally a state-writing governance problem. Once user content is committed to durable memory, safety must govern what agents write, not only what they say. PASB identifies the write-time controls needed to gate risky commits while preserving the source, role, and scope of stored content beyond response-level mitigations.

Graph Feedback Controls Consensus and Clique Formation in Open-Weight Language-Model Populations

Multi-agent language-model (LM) systems often determine which agents communicate, yet routing is usually treated as an implementation detail. We ask whether routing itself determines whether a population converges on a shared convention or fragments into persistent cliques. We study open-weight agents spanning 1.1B-32B parameters in a controlled naming game, tracking both emitted labels and full first-token preference distributions over the allowed labels. Similarity-based routing can isolate emerging conventions and sustain fragmentation even when every agent interacts in every round. Matched controls show that this effect is not explained solely by uneven participation or model-family-specific score preferences: random rematching and policies that connect disagreeing groups improve coordination when partner-label history is retained, but not when it is absent. Exposure alone is nevertheless insufficient, as some mixed-model populations remain divided despite frequent cross-family interaction, although the same models coordinate homogeneously. Trajectory and controlled-history analyses further distinguish reaching consensus from maintaining it. Finally, ARC-Challenge and MMLU experiments show that routing changes how correct and incorrect answers propagate without reliably improving accuracy. These results establish the runtime interaction graph as a causal design variable whose effects depend jointly on memory, model response, and population composition.

ReasFlow: Assisting Reasoning-Centric Scientific Discovery in Applied Mathematics via a Knowledge-Based Multi-Agent System

Recent advances in Large Language Models have fueled autonomous AI agents capable of tackling complex scientific tasks, yet existing automated research systems remain predominantly focused on empirically driven domains with quantitative benchmarks, leaving theory-driven discovery, particularly in mathematically grounded disciplines requiring rigorous proofs and synthesis of domain knowledge, largely underexplored. Key challenges include the difficulty of verifying theoretical reasoning at scale, insufficient reasoning ability for autonomous frontier exploration, and a scarcity of procedural heuristics in the literature. We introduce ReasFlow, an end-to-end autonomous agent system for reasoning-centric scientific discovery that operationalizes a collaborative paradigm where the human expert acts as Principal Investigator while the agent executes rigorous derivations as a capable graduate student. ReasFlow incorporates (i) a robust internal verification loop that audits logical coherence and corrects fundamental errors prior to human inspection, and (ii) an automated knowledge retrieval and self-improvement mechanism that proactively surfaces both declarative facts and overlooked procedural heuristics, substantially reducing expert intervention. The system unifies literature synthesis, algorithm design, theorem proving, experimentation, and manuscript preparation in a single system. Deployed to autonomously generate five complete research papers with rigorous theoretical and empirical content from minimal prompts, ReasFlow consistently achieves the highest evaluation scores among state-of-the-art open-access baselines under a curated LLM-based review rubric. ReasFlow is publicly accessible via the ReasLab platform, providing a collaborative workspace for AI-assisted theoretical research. Github repo: https://github.com/reaslab/ReasFlow.git.

MedFailBench: A Clinician-Built Open-Source Benchmark for Medical AI Safety Boundary Inspection

Most medical AI benchmarks measure whether a model knows the correct answer. MedFailBench asks a different question: which safety boundary failed? We present a synthetic benchmark and failure atlas built by a clinician. The resource labels medical AI errors by severity from 1 to 5 and safety gate type: missed urgent escalation, unsafe remote dosing, unsafe discharge reassurance, evidence fabrication, unsafe protocol execution, and source support gap. The current public release (v0.2.1) contains 44 synthetic cases reviewed by a clinician, with severity annotations, a public Hugging Face Space source, a safety gate taxonomy, a clinical severity rubric, and an automated pipeline for archiving model response screening runs. Forty cases have a populated safety gate field, and four require gate completion. No patient data, clinical validation claims, or model rankings are included. MedFailBench is released under Apache-2.0 and CC-BY-4.0 and carries the Zenodo DOI 10.5281/zenodo.21205535.

Knowledge-Centric Agents for Workflow Generation in ComfyUI

Workflow generation in visual creation systems such as ComfyUI demands not only syntactic accuracy but also expert-level reasoning over modular compositions. Existing large language model (LLM) approaches often treat this as a direct text-to-JSON generation task, struggling with structural brittleness and lacking the experiential knowledge required for effective design. We argue that successful workflow generation requires modeling knowledge itself, including its structure, hierarchy, and reasoning dynamics. To this end, we propose a knowledge-centric framework that learns to invert, inject, and infer with knowledge across multiple abstraction levels. We first perform knowledge inversion to distill hierarchical representations, ranging from full pseudo-codes and skeletons to high-level strategies, from large collections of real-world workflows. We then conduct knowledge injection through supervised fine-tuning, teaching the model to reason from task descriptions to strategies and from strategies to executable structures. During inference, the model performs reversible reasoning to synthesize executable workflows, augmented by self-refinement for structural coherence. Extensive experiments demonstrate that our method produces workflows with richer node diversity, more coherent structures, and higher execution success rates than existing systems, establishing a new foundation for knowledge-driven, agentic workflow generation.

Position: AI/ML Deepfake Research is Misaligned with AI-Generated Non-Consensual Intimate Imagery (AIG-NCII)

AI-generated non-consensual intimate imagery (AIG-NCII) is not adequately addressed in AI/ML literature regarding AI-generated media, commonly referred to as "deepfakes". While research on deepfakes currently focuses on its epistemic harms -- or harms relating to truth and authenticity -- this is misaligned with the dominant reality of generative AI abuse involving sexualized imagery. We conduct a landscape analysis of highly-cited works to demonstrate that technical interventions addressing deepfakes almost entirely ignore AIG-NCII, limiting the research ecosystem to authenticity detection tools. In this position paper, we argue that existing interventions address viewer-centric epistemic harms, such as fraud or scams, but ignore subject-centric dignity harms, such as AIG-NCII. We illustrate that knowing an image is synthetic does not mitigate harms to subjects and may, in some cases, even exacerbate them. We conclude by offering recommendations to realign the field, including updating threat models to consider subject-centric harms and addressing AIG-NCII in AI safety research. Finally, we caution that researchers should only engage in this high-risk domain if they implement safety guardrails for both subjects and researchers and establish partnerships with domain experts in sexual violence prevention.

Neuro-Symbolic Meta-Policies for Temporal Knowledge-Graph Memory under Partial Observability

Partially observable reinforcement learning requires deciding what to retain, retrieve, and forget over time. We introduce a neuro-symbolic meta-policy that learns which symbolic memory heuristic to apply at each decision point while keeping execution symbolic. Our setting uses temporal knowledge-graph memory in RoomKG, where hidden state and observations are represented as Resource Description Framework (RDF) graphs and memory is augmented with temporal RDF triple annotations. The model combines knowledge-graph encoding of memory contents with value heads for question answering, exploration, and forgetting, yielding a controller that is both adaptive and inspectable. This gives the work a direct Semantic Web grounding through RDF-based representation, annotation-compatible graph semantics, and graph-based symbolic operations over explicit memory state. On train/test room splits at long-term memory capacity of 512, the qualifier-aware StarE-GNN configuration achieves the best held-out performance among the compared symbolic, neural, and neuro-symbolic systems while preserving step-level traceability of memory-management decisions.

Athena-Brain Technical Report: An Efficient Robot Brain for General Intelligence and Embodied Interaction

Large language models (LLMs) have demonstrated remarkable capabilities in language understanding, reasoning, and world knowledge. As embodied agents become increasingly capable, there is a growing demand for compact models that can serve as an on-device brain, preserving the broad general intelligence of LLMs while enabling effective high-level interaction with embodied environments. Existing approaches, however, often prioritize either general-purpose intelligence or specialized embodied capabilities, making it challenging to satisfy both requirements within a single model. We present \textbf{Athena-Brain-8B}, an 8B LLM designed to serve as an on-device brain for embodied intelligence for embodied intelligence. Through a multi-stage post-training pipeline consisting of General Supervised Fine-Tuning, General Reinforcement Learning, Embodied Expert training, and Model Merge, Athena-Brain-8B maintains strong general capabilities while acquiring strong high-level embodied interaction capabilities and generating concise responses for efficient embodied interaction. Experimental results demonstrate the effectiveness of Athena across both general and embodied evaluations. Compared with the corresponding Qwen3-8B thinking model, Athena-Brain-8B achieves comparable performance on general language and reasoning benchmarks while generating substantially shorter responses. On in-domain embodied benchmarks, Athena-Brain-8B consistently outperforms models of similar scale and surpasses several substantially larger frontier models evaluated zero-shot, demonstrating that compact language models can effectively integrate strong general intelligence with embodied capabilities.

Quality Action Assurance: Multimodal Verification of Examiner Claims in VR OSCEs

Objective Structured Clinical Examinations (OSCEs) are the gold standard for assessing clinical competence, yet scoring remains vulnerable to examiner subjectivity, fatigue, and cognitive bias. Standard examiner validation via inter-rater statistics lacks explanatory power regarding the source of errors, as it neither analyzes examiner reasoning nor verifies examiner claims against actual events. Thus, we introduce Quality Action Assurance (QAA), a multimodal framework that verifies examiner claims in Virtual Reality (VR) pediatric OSCEs by comparing actions claimed by examiners against the true sequence of events, constructed from video, VR logs, and actor data. QAA combines a constrained temporal action alignment model, which performs action localization and actor source attribution, with a large language model that extracts examiner claims and checks them against the record. Across a 5-fold cross-validation, QAA achieves 99.2% $\pm$ 0.7% Actor F1 and 93.4% $\pm$ 1.9% W@16 for temporal alignment. Overall, QAA detects examiner errors with 69.9% precision and 76.7% recall, improving factual correctness from 39.2% to 79.2%, enabling fairer OSCE assessment.

AdaRoPE: Not All Attention Heads Should Rotate and Scale Equally

Rotary Position Embedding (RoPE) is widely adopted in Transformers to encode positional information, yet standard implementations enforce a uniform frequency schedule and scaling across all attention heads. Using simplified retrieval tasks and length generalization scenarios, we show -- both empirically and theoretically -- that heads with different functional roles require distinct frequency ranges and attention scaling factors to operate effectively. Ignoring this structure leads to suboptimal utilization of embedding dimensions and degraded performance, particularly under long-context settings. To address these limitations, we propose AdaRoPE, which equips each attention head with learnable rotation frequencies and attention scaling factors. Pretrained LLMs with AdaRoPE consistently outperform existing RoPE variants, including partial RoPE and NoPE baselines. For context extension, we further show that uniform frequency and attention scaling, used in methods such as YaRN, are suboptimal. By applying head-specific scaling, AdaRoPE enables better context extension while better preserving short-context performance in both the extrapolation setting and the long-context continued pretraining setting. These results highlight the importance of optimizing rotary position embedding at the level of individual attention heads.

Efficient and Interpretable Body-Based Emotion Recognition with Lightweight Temporal Convolutional Networks

Body-based emotion recognition is important for real-time affective systems, but graph-based skeleton models can be computationally expensive. This paper studies whether lightweight temporal convolutional networks (TCNs) can provide an efficient and interpretable alternative for body-based emotion classification. We evaluate a family of TCN models on DIEM-A and compare them with a graph-based time-series graph (G-TSG) baseline using accuracy, macro-F1, parameter count, and inference latency. Although G-TSG achieves the highest mean performance, TCN-Base remains within $1.58$ accuracy points and $1.25$ macro-F1 points while using $79.18\%$ fewer parameters and reducing classifier latency by approximately $12.5\times$. We also analyze body-region contributions using region-specific TCN models, zero-based occlusion, and G-TSG gradient saliency. The results show that upper-body motion provides the strongest standalone regional cue, that the usefulness of body regions varies across emotions, and that different interpretability methods capture distinct aspects of model behavior. These findings suggest that lightweight TCNs can support efficient body-based emotion recognition while also providing practical insight into how motion cues contribute to classification.

EmoAgent-R1: Towards Multimodal Emotion Understanding with Reinforcement Learning-based Dynamic Agent Specialization

Multimodal large language models (MLLMs) have achieved impressive performance in multimodal emotion recognition (MER) tasks and lifted MER to a new level that is complex emotion understanding with advanced video understanding abilities and natural language description. However, existing MLLM-based methods often use a fixed prompt to perceive the emotions, ignoring the dynamicity and complexity of the emotion source in the multimodal inputs. To address these issues, we propose a novel Reinforcement Learning-based Dynamic Agent Specialization framework (\textbf{EmoAgent-R1}) to optimize the emotion recognition, reasoning, and generalization abilities of an MLLM with dynamic agent specialization based on reinforcement learning. Specifically, we first adopt a cold start strategy to endow an MLLM with preliminary emotion recognition, reasoning, and agent routing ability by training with synthetic answer-conditioned chain-of-thought data and agent routing data. Then, we further train the MLLM with reinforcement learning to perceive emotions in a two-step agentic workflow with agent selection and agent specialization. To effectively train EmoAgent-R1, we propose a novel Progressive Group-Relative Policy Optimization (P-GRPO) to combine group-based relative advantages with a PMI-inspired progressive token-level modulation to transform sparse rewards into fine-grained learning signals, mitigating the coarse-grained uniform credit assignment issue in GRPO. Extensive experiments on MER benchmarks demonstrate the superiority of our EmoAgent-R1 in stronger emotion reasoning performance and improved optimization stability.

V-DEAL: Diagnosing Video Safety De-Calibration as an Understanding-Refusal Coupling Failure

As Video Large Language Models are increasingly deployed in real-world applications, ensuring their safety alignment has become critical. Counterintuitively, we find that harmful videos paired with benign queries achieve higher attack success rates than the same videos paired with explicitly harmful queries. To understand the underlying mechanism of this vulnerability, we present V-DEAL, a three-level diagnostic framework that jointly analyzes this failure across model behaviour, understanding, and internal representations. By progressively ruling out perception failure and quantifying the model's internal refusal tendency, V-DEAL provides a new diagnostic perspective for analyzing the underlying mechanism of the observed vulnerability. We tested six Video LLMs on three public benchmarks and observed that models correctly recognize harmful video content with over 81\% accuracy, yet the average attack success rate still reaches 48.33\% under the condition pairing harmful videos with benign queries. Hidden-state analysis further shows that visual understanding activates a weaker refusal tendency than textual understanding. Furthermore, we introduce a prompt injection intervention method that reduces attack success rates by an average of 48.24 percentage points and achieves performance comparable to prior fine-tuning-based methods, providing an effective and practical means to address such safety risks in Video LLMs.

The Optimal Sample Complexity of Linear Contracts

In this paper, we settle the problem of learning optimal linear contracts from data in the offline setting, where agent types are drawn from an unknown distribution and the principal's goal is to design a contract that maximizes her expected utility. Specifically, our analysis shows that the simple Empirical Utility Maximization (EUM) algorithm yields an $\varepsilon$-approximation of the optimal linear contract with probability at least $1-\delta$, using just $O(\ln(1/\delta) / \varepsilon^2)$ samples. This result improves upon previously known bounds and matches a lower bound from D\"utting et al. 2025 up to constant factors, thereby proving its optimality. Furthermore, our result establishes the stronger guarantee of uniform convergence: the empirical utility of every linear contract is an $\varepsilon$-approximation of its true expectation with probability at least $1-\delta$, using the same optimal $O(\ln(1/\delta) / \varepsilon^2)$ sample complexity.