AI News

Reconstruction differences rival individual differences in the fly compass connectome and break trained circuit models (arxiv.org)

A Unified Evaluation Framework of Physical Adversarial Robustness in the Edge VLM Era (arxiv.org)

A Decade of Learning-Based Control: From Theoretic Foundations to Practical Deployment (arxiv.org)

TeachBack: Learning to Teach via Bidirectional Preference Optimization for On-Policy Self-Distillation (arxiv.org)

From Augmentation to Synthetic Evidence: Evaluating Generated Chest Radiographs at Dataset Intake (arxiv.org)

SAFTER: Safety-Aware Filtering Through Explicit Recoverability for Robust Reinforcement Learning under Distribution Shift (arxiv.org)

First Passage Attention (arxiv.org)

Breaking Diffusion: Parallel Reasoning with Draft Refinement (arxiv.org)

Self-normalized Bernstein Inequalities for Regenerative Processes and Applications to Learning Without Independence (arxiv.org)

Velocity Ambiguity Profiles: Time-Resolved Bayes-Risk Diagnostics for Flow Matching (arxiv.org)

BatchBench: Benchmarking AI Robustness to Batch Effects and Domain Shift in Biomedical Images - A JUMP Cell Painting Leaderboard (arxiv.org)

Differentially Private Sparse Reward Estimation with Preference Feedback (arxiv.org)

Orbitol: Crystal Generation on Wyckoff Orbits (arxiv.org)

Circuits of LLM Classifiers (arxiv.org)

From Classifier Heads to Encoder Prototypes: Personalized Federated Learning with Head-Derived Clustering (arxiv.org)

Loopholing All the Way: Language Modeling with pure Self-Conditioning (arxiv.org)

ExGrad: Beyond Suppression - Explicit Clean-Gradient Subtraction in Adversarial Training (arxiv.org)

RubRIX: Rubric-Driven Risk Mitigation in Caregiver-AI Interactions (arxiv.org)

Synthesis and Verification of Register Automata over Data Sequences (Technical Report) (arxiv.org)

Learning to Interfere in Non-Orthogonal Multiple-Access Joint Source-Channel Coding (arxiv.org)

AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems (arxiv.org)

Bilateral Sharpness-Aware Minimization for Flatter Minima (arxiv.org)

Copying Under Repetition: From Prefix Matching to Position Hashing (arxiv.org)

Self-organizing reinforcement learning: a learner-free swarm approaches the explore-exploit optimum (arxiv.org)

TrustAgent-RFF: Agentic Diagnostic Planning for Reliable RF Fingerprinting under Domain Shift (arxiv.org)

An important Bottleneck in LLM Optimizers: Principle-Level Local Optima (arxiv.org)

Compression Through Time: An Interplay of Separability, Stationarity and Capacity (arxiv.org)

TrustmeWatcher: An Application for Workplace Micro-Sensing and Explainable Well-Being Feedback (arxiv.org)

Physics is the Best Teacher: Consistency Learning for Time-Invariant Operators of Chaotic Dynamics (arxiv.org)

A Generalisation Signal Need Not Be a Model-Selection Signal (arxiv.org)

Tractable Inference with Kolmogorov-Arnold Networks (arxiv.org)

RangeNav: Task-Aligned LiDAR Range-Image Pretraining for Robot Navigation (arxiv.org)

NeuroCodec: EEG-Guided Latent-Domain Generative Model for High-Fidelity Target Speaker Extraction (arxiv.org)

Real PMIDs, Wrong Papers: Harness Design for Biomedical Claim-Verification Agents (arxiv.org)

Where Do LLMs Learn to Recommend AI? (arxiv.org)

Rethinking Post-hoc Debiasing of Large Pretrained Models (arxiv.org)

Supermodular Rank: Set Function Decomposition and Optimization (arxiv.org)

TraceForge: Distilling with Your Own Thinking (arxiv.org)

Lineage-Aware Search Control for LLM-Based Evolutionary Factor Mining (arxiv.org)

Recursive Code Evolution for Task-Oriented Semantic Communications (arxiv.org)

Space weather dependence of decameter-scale ionospheric irregularities spatial distribution (arxiv.org)

Watch Your Step: Anticipated Perception for Humanoid Locomotion (arxiv.org)

No Frames, Just Frequencies: Bandwidth-Efficient Deployment-Time Adaptation for Smart Glasses (arxiv.org)

SCDM: Scalable Causal Discovery in Nonlinear Temporal Systems with Meta-Learning (arxiv.org)

CORD: Conflict-Aware Optimization for Joint RL and Distillation in Agents (arxiv.org)

FrontierCS 2: A Benchmark for Computer Science R&D (arxiv.org)

Your Benchmark Is an Empirical Measure Over Difficulty (arxiv.org)

A Sharp Dynamic Haussler Lemma (arxiv.org)

Monitorability Is Trainable: Post-Training for Chain-of-Thought Monitorability (arxiv.org)

Structure-Aware Message Passing on Petri Nets through Directed Walks and Incidence Diffusion (arxiv.org)