AI News

MUSE: A Unified Agentic Harness for MLLMs (arxiv.org)

COD10K-C: Benchmarking Robustness of Camouflaged Object Detection Under Natural Image Corruptions (arxiv.org)

q0: Primitives for Hyper-Epoch Pretraining (arxiv.org)

From Answers to States: Verifiable Process-Level Evaluation of Chemical Reasoning in Large Language Models (arxiv.org)

Diagnosing Knowledge Gaps in LLM Tool Use: An Agentic Benchmark for Novel API Acquisition (arxiv.org)

TiWeaver: Unified Temporal Dynamics Modeling via Contextual Patching (arxiv.org)

Position: Prioritize Identifying Structure, Not Complex Models, for Scientific Discovery (arxiv.org)

EvoDrive: Pareto Evolution for Safety-Critical Autonomous Driving via Self-Improving LLM Agents (arxiv.org)

Safety Measurements for Fine-tuned LLMs Should be Grounded in Capability (arxiv.org)

TurtleAI: Benchmarking Multimodal Models for Visual Programming in Turtle Graphics (arxiv.org)

DDOR: Delta Debugging for Explainable Overrefusal Testing and Repair (arxiv.org)

The Epi-LLM Framework: probing LLM behavioral priors through epidemiological agent-based models (arxiv.org)

CoEval: Ranking Language Models for Custom Tasks Without Labeled Data or Trustworthy Benchmarks (arxiv.org)

Black-box, Adaptive, Efficient, Transferable, Harmful, Applicable... Attacks Are All You Need to Break LLMs (arxiv.org)

Qwen-Image-Flash: Beyond Objective Design (arxiv.org)

Resource-Constrained Adaptive Inference for Sequential Pricing (arxiv.org)

Target Updates May Stabilize Linear Q-Learning: Periodic and Soft Dynamics (arxiv.org)

Training-Free Multi-Concept LoRA Composition with Prompt-Aware Weighting (arxiv.org)

AI Agents Enable Adaptive Computer Worms (arxiv.org)

Gender-Dependent Diagnostic Substitution in LLM Medical Triage: Same Symptoms, Unequal Urgency (arxiv.org)

State-Coupled Volatility in Latent Dynamical Systems: Recovery Under Partial Observation (arxiv.org)

Synthetic Hallucinations, Real Gains: Hard Negatives from Frontier Models for FIM Hallucination Mitigation (arxiv.org)

DiffUNet^2: Bidirectional Prediction, Probabilistic Generation and Collaborative Visual Discovery for Scientific Data (arxiv.org)

Merit or networks? What decides where research is published (arxiv.org)

SEAOTTER: Sensor Embedded Autoencoding with One-Time Transcode for Efficient Reconstruction (arxiv.org)

AugMask: Training Diffusion Models on Incomplete Tabular Data via Stochastic Augmentation and Masking (arxiv.org)

MLSkip: Data Skipping for ML Filters via Lightweight Metadata (arxiv.org)

Finding Needles in the Haystack: Transductive Active Labeling in Ecology (arxiv.org)

PINNfluence: Interpreting PINNs through Influence Functions (arxiv.org)

Denoise First, Orthogonalize Later: Understanding Momentum in Muon via Spectral Filtering (arxiv.org)

Building Trust in Black-box Optimization: A Comprehensive Framework for Explainability (arxiv.org)

Easy-to-Use Shielding for Reinforcement Learning (arxiv.org)

R2DN: Scalable Parameterization of Contracting and Lipschitz Recurrent Deep Networks (arxiv.org)

A Geometric Lens on Physics-Aligned Data Compression (arxiv.org)

PHASE: Physiology-Aware Hyperspectral Reconstruction via Object-to-Human Domain Adaptation (arxiv.org)

Towards Non-Monotonic Entailment in Propositional Defeasible Standpoint Logic (arxiv.org)

Code-on-Graph: Iterative Programmatic Reasoning via Large Language Models on Knowledge Graphs (arxiv.org)

HARVE: Hacking-Aware Reward-Head Vector Editing for Robust Reward Models (arxiv.org)

Multi-Modal Machine Learning for Breast Cancer Recurrence Prediction (arxiv.org)

Rethinking Neural Width for Alternating Current Optimal Power Flow Proxies (arxiv.org)

InftyThink+: Effective and Efficient Infinite-Horizon Reasoning via Reinforcement Learning (arxiv.org)

FGRPO: Federated GRPO with Adaptive Aggregation on Non-IID Data (arxiv.org)

Let There Be Light: Reflection, Refraction and Scattering for Neural Operators (arxiv.org)

Calibrating Urban Traffic Simulation from Sparse Road Observations via Genetic Optimization (arxiv.org)

Proof-Refactor: Refactoring Generated Formal Proofs into Modular Artifacts (arxiv.org)

EqGINO: Equivariant Geometry-Informed Fourier Neural Operators for 3D PDEs (arxiv.org)

Learning Temporal Causal Structure via Smooth Differentiable Optimization (arxiv.org)

Constitutional On-Policy Safe Distillation (arxiv.org)

SegTune: Structured and Fine-Grained Control for Song Generation (arxiv.org)

Calibration Data Trade-offs Across Capability Dimensions: Why Multi-Source Mixing Matters for High-Sparsity LLM Pruning (arxiv.org)