AI News

When Pleasing the User Compromises Care: AI Sycophancy in Caregiver-Mediated Healthcare (arxiv.org)

ACED-Bench: Evaluating How LLM Agents Acquire and Act on Causal Evidence (arxiv.org)

ANASSA: From High-Level Framework to Low-Level Architecture for Spatial Agentic AI (arxiv.org)

Variance-Optimal State Resampling for Reinforcement Learning with Verifiable Rewards (arxiv.org)

Measuring the Inference Energy Efficiency of Open-Weight LLMs across Benchmarks, Reasoning Modes, and Hardware (arxiv.org)

SAT-RL: Training Language Models for Logical Reasoning with Symbolic Solvers (arxiv.org)

Frustratingly Easy Agentic RAG via Portable Skills: A Case Study in Customer Support (arxiv.org)

Energy Ring: Stabilizing Feature Energy for Reliable Test-Time Discovery (arxiv.org)

AlignPoison: Conditional Targeted Poisoning of Generative Recommenders via Decoding-Aligned Prefixes (arxiv.org)

Adaptive Computation with Conformal Certification (arxiv.org)

Spectral-MoE: Heterogeneous Per-Layer Expert Utilization in Dense-to-MoE Conversion (arxiv.org)

LLMs program using Reusable and Composable Abstract Representations (arxiv.org)

Tracing Morphology, Semantics, and Divergence in Language Acquisition (arxiv.org)

MvFFN: Multi-view Floor plan Feed-forward Network for Unposed Wide-Baseline Panorama Layout Reconstruction (arxiv.org)

HiPPT: Physiology-Grounded Wearable PPG Representation Learning via Multi-Scope Text Alignment (arxiv.org)

Reels-RSI: Program-Level Self-Improvement for Code-Rendered Explainer Videos (arxiv.org)

MontageFlow: Physically Valid High Fidelity EEG Generation (arxiv.org)

ScienceBuddy-PV: A Verifier-Governed Recursive Improvement Protocol for Panvascular Intervention Research (arxiv.org)

FDE-Bench: Evaluating End-to-End Delivery from Underspecified Real-World Business Requests (arxiv.org)

Randomly Wrong Signals: Bayesian Auction Design with ML Predictions (arxiv.org)

Natural Code Transformation Confuses Agents in Vulnerability Analysis (arxiv.org)

Strong Gaussian Approximation of Constant Learning Rate SGD: Bias and Trajectory Inference (arxiv.org)

AlphaRLM: Aligning Language Models for Literature-Grounded Symbolic Factor Discovery (arxiv.org)

SuperAI: Merging Capabilities Across Heterogeneous Foundation Models (arxiv.org)

Where to Start, How to Move: Analyzing Curricula under GRPO for Long-Horizon LLM Agents (arxiv.org)

Are Multimodal Benchmarks Really Useful? Item-Level Multimodal Benchmark Diagnosis via Structure-Response Co-Calibration (arxiv.org)

PsyMemBehav: Psychology-Informed and Memory-Grounded Two-Level Identity-to-Action Reasoning for Social-Media Behavior Simulation (arxiv.org)

Learning after Execution with Action-Specific Credit (arxiv.org)

CB-Agent: Concept-Based In-Context Learning for Sequential Action Selection (arxiv.org)

Fully sequential Gaussian process regression for preference learning (arxiv.org)

Iterate on an LLM, Serve on an SLM (arxiv.org)

Decoupling Knowledge Forgetting and Refusal Learning in MLLM Unlearning (arxiv.org)

From Existing Methods to New Research Ideas: Two Perspectives on Recursive Self-Improvement (arxiv.org)

PEARL: Process-Experience Aware Reinforcement Learning for Multi-Turn Agents (arxiv.org)

Energy-Based Models For Speech Synthesis (arxiv.org)

AegisVox: Factorized Privacy and Clinical Preservation for Long-Form Speech (arxiv.org)

Shaping Latent Mixture-of-Experts: From Expert Geometry to Layerwise Capacity Allocation (arxiv.org)

RASSG: Repair-Aware Semantic Survival Gating for Streaming Tool Execution (arxiv.org)

A Hybrid Kinematic GCN and Elastic Transformer for Efficient Indian Sign Language Recognition (arxiv.org)

AcrossVAM2.0: Conflict-Aware Point Supervision for Parameter-Efficient Video Diffusion Adaptation (arxiv.org)

Joint Discrete and SE(3) Diffusion for Protein Interface Design (arxiv.org)

DSABench: Benchmarking Source Discovery, Coordination, and Execution for Multi-Database Agents (arxiv.org)

How to Build a Benchmark That Actually Measures Taste (arxiv.org)

Relational Semantic Grounding for Test-Time Discovery (arxiv.org)

MHWA: Multi-timescale Hierarchical World-Action Model (arxiv.org)

Think in Strategies, Act with Tools: Traceable Agentic Optimization for Combinatorial Problems (arxiv.org)

SAGE: Pushing 2-Bit Vector-Quantized LLMs toward Real-World Usability (arxiv.org)

Equal-information evaluation of credit assignment in modular financial reinforcement learning (arxiv.org)

GateSight: Recurrent Memory Anticipates Attention Memory for Query-Agnostic KV Eviction in Hybrid LLMs (arxiv.org)

CUSE: Fine-Grained Multimodal Unlearning with Counterfactual Guidance and On-Policy Self-Distillation (arxiv.org)