AI News

MedEvo: Medical Reasoning Evolution through Knowledge Navigation (arxiv.org)

Stable Token-level Scheduling for Serverless LLM Inference (arxiv.org)

ECG-RSI: Recursive Self-Improvement through Targeted Electrocardiogram Waveform Evolution (arxiv.org)

Encounter-Driven City-Scale Human Mobility Generation with Large Language Models (arxiv.org)

Cowork-Gym: Synthesizing Interactive Work Environments at Scale (arxiv.org)

RoTA-Bench: Evaluating 3D Spatial Reasoning through Interactive Projection Alignment (arxiv.org)

Can Agents Truly Forget? Explore Machine Unlearning for LLM Agents (arxiv.org)

When Should Enterprise Policy Decisions Be Learned, and When Should They Be Reasoned? (arxiv.org)

Beyond Bias Scores: Answerability Confounds in Activation Steering with Gradient-Constrained Direction Design (GCD) (arxiv.org)

Executable Prefix Values for Locating Agent Failures (arxiv.org)

MAS-Gate: Unlocking Multi-Agent System Lifecycle via Node-Centric Quantification (arxiv.org)

OpenDepth: Repurposing Image Editing Priors for Robust Open-World Monocular Depth Estimation (arxiv.org)

Verifier-Harness Co-Evolution for Training-Free Self-Improvement on Open-Ended Tasks (arxiv.org)

RAN-JEPA: An Action-Conditioned World Model for Energy-efficient Mobile Network Control (arxiv.org)

TacitVLA: Reasoning Vision-Language-Action Models Without Reasoning Latency (arxiv.org)

REA: From Correct LLM Judgments to Safe Decision Delivery (arxiv.org)

An FPGA-Based Spiking Neural Network for Spatiotemporal Pattern Learning and Classification (arxiv.org)

Large Language Models Extend Known Research But Do Not Create New Research Directions (arxiv.org)

Reading Right, Acting Wrong: Diagnosing Decision Failures of System One Models in Agent Loops (arxiv.org)

SchemaWeaver: Accelerating On-Device Mobile Agentic AI via Schema-Guided Parallel LLM Decoding (arxiv.org)

Agency as a New Organizing Abstraction for Computer Science (arxiv.org)

Generalizable field of view extension for diffusion MRI via conditional flow matching (arxiv.org)

Learning to Play: Reinforcement Learning and Games (arxiv.org)

EPOCH: Coverage-Aware Evolution with Experiential Preference Optimization for LLM-based Automated Heuristic Design (arxiv.org)

Get In, Decide, Prove: A Merchant-Side Readiness Benchmark for Agentic Commerce (arxiv.org)

What Makes a Latent Space Shared? Direct and Recoverable Compatibility in Motion VAEs (arxiv.org)

When Goals Leak Across Relationships: How Cached Goal Representations Shape Cross-Client Decisions (arxiv.org)

Beyond Black Boxes: Visual Grounding for Text Embeddings (arxiv.org)

Budgeted KV-Cache Repair for Evolving Contexts (arxiv.org)

Do Theorem Provers Need Theorem-Proving Agents? mini-Lean-agent: A Coding-Agent Baseline for Lean (arxiv.org)

Beyond Successful Play: Diagnosing Opponent Modeling in Large Language Models (arxiv.org)

Reliable Text-to-SQL in Resource-Constrained Settings with Small LLMs via Iterative Retry Workflows (arxiv.org)

The Agent Knew: Internal States Forecast Unauthorized Actions and Stop Them, and the Lies That Follow (arxiv.org)

Towards Resilience: Slice-Driven Orchestration of Mobile IoMT in 6G-Edge-Continuum (arxiv.org)

PhysCodeGym: Benchmarking Coding Agents for Physically Verifiable 4D Reconstruction (arxiv.org)

RL Post-Training Diffusion Language Models Along Their Own Generation Paths (arxiv.org)

Logit Leakage through Floating-Point Bias Addition (arxiv.org)

HASTE: Hierarchical and Attribute-aware Supervision for Generative Recommendation (arxiv.org)

claimcheck: Measuring and Improving Specification Faithfulness (arxiv.org)

Reasoning for Propaganda Detection via Label-Guided CoT Optimisation (arxiv.org)

On the Statistical Mechanics of Attention (arxiv.org)

Thinking Outside the Bots: Human Ideas in AI's Semantic Blind Spots Are Not Inferior (arxiv.org)

Joint Embedding with Dynamic Integration network (JEDI-net) (arxiv.org)

StartLux-Decision: AutoThink Decision Models with Reinforcement Learning for Calibrated Deliberative Decisions (arxiv.org)

MID: Mask-Image Distributional Divergence for Evaluating Medical Image Segmentation (arxiv.org)

Phase-Conditioned Affordance Fields for SE(3)-Equivariant Visuomotor Policies (arxiv.org)

Measuring Learned Equivariance: An Information Theoretic Approach (arxiv.org)

From Conceptual EneMonIE to an Executable Backend: A Modular Architecture for Energy Information Extractionand Grounded Question Answering (arxiv.org)

Compute-optimal data scaling for neural surrogates via multi-fidelity training (arxiv.org)

Deleting Code Without Losing Knowledge: Regenerative Software for Python Libraries (arxiv.org)