AI News

Narrative Paradoxes in LLM Inference: Shaping Reasoning Trajectories toward Misalignment (arxiv.org)

ReTAC: Residual Test-Time Agentic Computation for Video Spatial Reasoning (arxiv.org)

HyWe: Unifying Search and Operation for Real-World Hybrid Web Tasks (arxiv.org)

Context Value Informed In-Context Reinforcement Learning (arxiv.org)

Efficient Test-Time Adaptation For Robot Policies (arxiv.org)

FOLD: Position-Free Far Memory with Local Order for Long-Context Speculative Decoding (arxiv.org)

Cortex: Consolidating Context into Parametric Memory (arxiv.org)

HRTF Upsampling and Personalization via Neural Operators (arxiv.org)

What Does Disagreement Mean? A Semantic Identifiability Frontier for Pairwise Preference Learning (arxiv.org)

TRAP: Tool-result Replay Attack for Probing Misbinding in LLM Agents (arxiv.org)

Efficient Multi-Agent Reasoning via Confidence-Guided Adaptive Debate (arxiv.org)

ChronoScribe: Semantic-Numerical Specialization for Instruction-Driven Time Series Forecasting (arxiv.org)

Same Operation, Different Permission: Evaluating Whether Guardrails Honor Authorization (arxiv.org)

Generation of data on discontinuous manifolds via continuous stochastic non-invertible networks (arxiv.org)

The COUNT: Counting N-grams Unlocks Scaling on Host Memory and Training-Free Adaptation (arxiv.org)

SafeAdapt-Poker: Counterfactual Risk Calibration for Budgeted Adaptation under Opponent Shift (arxiv.org)

Long-Horizon RL Needs Muon (arxiv.org)

World Models Need a Physical Registry: Learning Internal Predictive States for Physically Plausible World Models (arxiv.org)

RetroPlanner: Multi-Route Expansion and Multi-Objective Reasoning for LLM-Based Retrosynthesis Planning (arxiv.org)

Recursive Environment Scaling for General Long-Horizon Agents (arxiv.org)

EvoWeb: A Benchmark for Coding Agents on Evolving Web Applications (arxiv.org)

Complex materials engineering via Provenance Action Graphs: potential, structure and demonstration (arxiv.org)

SLS-Bench: A Benchmark for Incident Log Summarization with Synthetic Observability Data (arxiv.org)

Chain Dimension Scaling: Communication-Efficient Loop-Style Scaling for Industrial Short-Video Ranking (arxiv.org)

The Anatomy of a Blind Spot: What an LLM Judge Must Be Shown, and How Much (arxiv.org)

VascuSR: Iterative Anatomy-Aware Diffusion for CT Super-Resolution (arxiv.org)

VASPO: Self-Distillation with Privileged Value Estimation (arxiv.org)

ActiveEvidence: Decision-Focused Evidence Graphs with Verifiable Acquisition Gaps (arxiv.org)

PrismaCoder: Pushing the Boundaries of Open Multimodal Code Intelligence through Decomposed Rubric Synthesis (arxiv.org)

Latent Compaction with Test-Time Training (arxiv.org)

Price of Anarchy of First-Price Auctions in Agentic Markets (arxiv.org)

Co-Evolving Acting and Forecasting for Language Agents (arxiv.org)

Modeling and Understanding Neural Weights of Humanoid Policies with Diffusion (arxiv.org)

Hierarchical Residual Policy Learning for Real-World Mobile Manipulation with Sparse Human Guidance (arxiv.org)

RAP-WAM: Learning World-Action Models through Predictive Rehearsal and Action Perturbations (arxiv.org)

XAI-Guided Fourier Domain Generalisation for Pure-Simulation Sim-to-Real Deep Reinforcement Learning (arxiv.org)

PXRD-Bench: How Close Are We to Determining Crystal Structures from Powder Diffraction? (arxiv.org)

Boosting the Learning for Ranking Patterns (arxiv.org)

EAU: Environment-Agent-User Co-evolution for Training User-Centric Agents (arxiv.org)

Rule-based and reinforcement-learning policies for resource allocation in a pediatric echo lab: a simulation study (arxiv.org)

Does the Explanation Drift Before the Guarantee? Label-Free Early Warning on Evolving Graphs (arxiv.org)

Overtaught, Underlearned: Learner-Centered Distillation from Multi-Agent Discussion (arxiv.org)

Making Language the Low-Rate Code of Vision: Successive Refinement from Words to Pixels (arxiv.org)

AI Discovery in Combinatorics: From Problems to Solutions and Back to Bigger Problems (arxiv.org)

A Versatile Variational Quantum Kernel Framework for Non-Trivial Classification (arxiv.org)

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

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

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

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

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