AI News

Fresh at Decision, Stale at Action: Routing New Information to LLM Decisions Already Made (arxiv.org)

SkillScriptBench: Benchmarking Self-Evolution of Executable Agent Skill Packages Beyond Markdown (arxiv.org)

Recursive Improvement of a Differentiable Scientific Software Ecosystem (arxiv.org)

Penumbra: Sample-Efficient Adversarial Search for Regulatory Obligations (arxiv.org)

PyINE: A Framework for Scalable Elicitation and Oversight via Code Execution (arxiv.org)

Do More Modalities Always Help? A Geometric Perspective on Missing-Modality Robustness (arxiv.org)

MS-Exam-Gen: An Automatic Benchmark Construction Toolkit for Evaluating LLMs on Multiple Sclerosis-MRI Imaging Knowledge (arxiv.org)

Evolving in Thought Space: Training a Small Model at Test Time Unlocks Better Discoveries (arxiv.org)

Hiking in the Wild: A Scalable Perceptive Parkour Framework for Humanoids (arxiv.org)

Open-MMUnlearning: Unifying Methods and Evaluation for MLLM Unlearning (arxiv.org)

Prediction Is All MoE Needs: Expert Load Distribution Goes from Fluctuating to Stabilizing (arxiv.org)

NeuSE: Neural SE(3)-Equivariant Embedding for Long-Term Object-Based Simultaneous Localization and Mapping (arxiv.org)

LLMRec: Benchmarking Large Language Models on Recommendation Task (arxiv.org)

Fast Diffusion-GAN Model for Generating Symbolic Music with Emotions (arxiv.org)

FlowWise: Stateful Fast-Slow Model Serving for Streaming Traffic Intelligence (arxiv.org)

Beyond Accuracy: Counterfactual Fragility and Demographic Bias in Clinical Evaluation of LLMs (arxiv.org)

GREAT: A Graph Edge Attention Network for the Electric Vehicle Routing Problem with Asymmetric Costs (arxiv.org)

Polyrating: A Cost-Effective and Bias-Aware Rating System for LLM Evaluation (arxiv.org)

Scale-specific auxiliary multi-task contrastive learning for deep liver vessel segmentation (arxiv.org)

Is Complex Query Answering Really Complex? (arxiv.org)

Open Foundation Models in Healthcare: Challenges, Paradoxes, and Opportunities with GenAI Driven Personalized Prescription (arxiv.org)

CompanionSafe: From Multi-turn Safety Evaluation to Guardrails for AI Companions (arxiv.org)

A Generalist Hanabi Agent (arxiv.org)

Horizon Scans can be accelerated using novel information retrieval and artificial intelligence tools (arxiv.org)

TopoPilot: Reliable Conversational Workflow Automation for Topological Data Analysis and Visualization (arxiv.org)

UnSAMv2: Self-Supervised Learning Enables Segment Anything at Any Granularity (arxiv.org)

Neural Evidence and Behavioural Welfare Economics: When Can Neuroeconomics Inform Policy? (arxiv.org)

Semantics as a Shield: Label Disguise Defense (LDD) against Prompt Injection in LLM Sentiment Classification (arxiv.org)

LifeSearch: Benchmarking Deep Research Agents for Real-World Life and Productivity (arxiv.org)

ECLIPTICA -- A Framework for Switchable LLM Alignment via CITA - Contrastive Instruction-Tuned Alignment (arxiv.org)

Less is More: Label-Guided Efficient Summarization of Procedural and Instructional Videos (arxiv.org)

NeuroVLA: A Brain-inspired Embodied Intelligence for Fluid and Fast Reflexive Robotics Control (arxiv.org)

Uni-RS: A Spatially Faithful Unified Understanding and Generation Model for Remote Sensing (arxiv.org)

PixSearch: Region-Grounded Retrieval for Knowledgeable Large Multimodal Models (arxiv.org)

Scalable Power Sampling: Unlocking Efficient, Training-Free Reasoning for LLMs via Distribution Sharpening (arxiv.org)

TTT-Parkour: Rapid Test-Time Training for Perceptive Robot Parkour (arxiv.org)

Flow Policy Gradients for Robot Control (arxiv.org)

DRMOT: A Dataset and Framework for RGBD Referring Multi-Object Tracking (arxiv.org)

Combining scEEG and PPG for reliable sleep staging (arxiv.org)

Articulated 3D Scene Graphs for Open-World Mobile Manipulation (arxiv.org)

Naga: Vedic Encoding for Deep State Space Models (arxiv.org)

When Agents Commit Too Soon: Diagnosing Premature Commitment in LLM Agents (arxiv.org)

Plans Fade Fast: Why Context Management Is Load-Bearing for LLM Agents (arxiv.org)

Litmus: Zero-Label, Code-Driven Metric Specification for Evaluating AI Systems (arxiv.org)

MIND: Lightweight and Effective Memory Injection Defense for LLM Agents via Intent-Aware Information Bottleneck (arxiv.org)

Toward Fine-Grained Forgetting:Attribute Unlearning for Multimodal Large Language Models (arxiv.org)

LaCache: Robust Semantic Caching for LLM Serving (arxiv.org)

CockpitHAT: Dependency-Graph-Driven Hierarchical Attribution for Embodied Multi-Agent Cockpits (arxiv.org)

SearchMaster: Grounded and Regulated Self-Play for Search Agents (arxiv.org)

LoopArena: Benchmarking Models as Runtime Controllers for Loop Engineering (arxiv.org)