AI News

Acceptable Misalignment (arxiv.org)

CLIVO: Differentiable Cluster-Based Instruction Mixing with Learned Validation Objectives (arxiv.org)

The Format is the Filter: Scoring Format Fixes What an LLM Judge Can Express (arxiv.org)

LIBERO-Industry: Benchmarking Robotic Manipulation under Industrial Process Constraints (arxiv.org)

Game-Creation-Agents: Autonomous Game Production (arxiv.org)

Geospatial Vector Geometry as a Native Spatial Modality for Vision-Language Models (arxiv.org)

Separating Prompt Harmfulness from Interaction Safety via Partial Generation (arxiv.org)

Lost in Translation: Why Lower Bias Scores Don't Mean Fairer Models (arxiv.org)

Parameter-Efficient Temporal Domain Generalization (arxiv.org)

Bridging Academia and Industry: A Comprehensive Benchmark for Attributed Graph Clustering (arxiv.org)

On the Dynamics of Delayed Generalization in Linear Models (arxiv.org)

Queue-Aware Token Communications: Bridging AI Representation and Wireless Service (arxiv.org)

Trace-to-Architecture: Capability Bottleneck Diagnosis for Harness Revision (arxiv.org)

Structural Baselines Matter for Agent Failure Detection and Attribution (arxiv.org)

Does a Hosted Decision Model Respect Irrelevant Alternatives? A Pre-Registered Test of Option-Set Dependence in Jev (arxiv.org)

LATCH: LLM-Assisted Timing Closure Harness for Post-Route FPGA Optimisation (arxiv.org)

Rational Transductors (arxiv.org)

Deep Time-series Forecasting Needs Kernelized Moment Balancing (arxiv.org)

SWE-Spot: Building Small Repo-Experts with Repository-Centric Learning (arxiv.org)

Interpolation of GEDI Biomass Estimates with Calibrated Uncertainty Quantification (arxiv.org)

Long-Range Distillation: Distilling 10,000 Years of Simulated Climate into Long Timestep AI Weather Models (arxiv.org)

When 1+1 Backfires: Rethinking Relevance-Diversity Balancing for Sample-Aware Visual Token Reduction (arxiv.org)

Double-Rollout Reinforcement Learning with Sampling-based Reward Estimation (arxiv.org)

G$^2$WAM: World Action Models for Efficient and Geometry Grounded Image-Goal Navigation (arxiv.org)

Audit tools of the trade: Automated Auditor Elicitation Can Be Sensitive to the Context the Auditor Is Given (arxiv.org)

AtomSearch: Evidence-Grounded Tree Search for Agent Skill Optimization (arxiv.org)

SAM3RL: Memory Control via Reinforcement Learning for Visual Object Tracking (arxiv.org)

Multimodal Medical Image Parsing with Task-Specific Low-Rank Adapters (arxiv.org)

Few-Shot Adaptation for Anomalous Sound Detection at Low False-Positive Rate (arxiv.org)

Explorer-Zero: Improving Self-Play for Long-Horizon Tasks with Decomposable Verifiable Rewards (arxiv.org)

GUIDE: Computationally Augmented Consensus for Trustworthy Clinical Practice Guidelines (arxiv.org)

The Sharp Directions Are Against You: Curvature Analysis of Activation Steering (arxiv.org)

Heterogeneous Agentic Coevolution (arxiv.org)

Beyond Single-Pathogen Forecasts: Predicting Shared Respiratory Hospital Burden from Influenza, COVID-19, and RSV (arxiv.org)

Hyperspherical Contrastive Representations Unlock Planning and Exploration (arxiv.org)

Training Language Models to Fail Gracefully (arxiv.org)

From Human Corrections to Learned Segmentations: A Critical Mathematical Review of Information, Learning, and Human Effort (arxiv.org)

ReAl: Refine neural operators via aligning dynamical flow (arxiv.org)

NeuroDETR: Multiscale Discrete Spatiotemporal Tokenization for Asynchronous EEG Event Detection (arxiv.org)

Understanding and Reducing Misalignment in Persuasion Optimization (arxiv.org)

Cross-Protocol Personality Assessment with LLMs: Aligning Behavioral Evidence with Human Ratings (arxiv.org)

Computational Reflective Equilibrium for Agentic AI: A Rawlsian Alignment Scaffold for Governance (arxiv.org)

Large language model-generated patterns for epidemiological modeling: a multi-benchmark evaluation framework (arxiv.org)

Handoff Curves: Initial Selection, Later Handoff, and Supervision Targets in Multi-turn LLM Routing (arxiv.org)

ScreenASR-Bench: A Tiered Benchmark for Speech Recognition with Visual Context (arxiv.org)

Revocation Is Not Complete at Receipt: An Executable Outcome Contract for Distributed Agent-Tool Gateways (arxiv.org)

PopGen: Generating Semantically Diverse Populations for Zero-Shot Coordination (arxiv.org)

Small but Mighty: Test Suites as Improvement Signals for Self-Evolving Agents (arxiv.org)

ADAPT: Audio-Depth Action Planning with Transition-Aware Rewards for Lights-Out Industrial Navigation (arxiv.org)

Vision: "What Don't I Know About User's Preferences?" How Uncertainty Estimation Improves Proactive Assistant Alignment (arxiv.org)