AI News

SkillFusion: Towards Reliable Multi-Skill Composition for LLM Agents (arxiv.org)

When Skills Remain But Access Fails: Factorized Skill Memory for Continual Skill Internalization (arxiv.org)

Spectral structure learning for clinical time series (arxiv.org)

L3M: A Verifiable Harness and Runtime for Self-Extending and Self-Improving Agents (arxiv.org)

A Graph-Based Analysis of Comparability, Scale, and Attribution in Private Transformer Inference (arxiv.org)

DAGFormer: Global Topology Prediction for Transformer Cross-Layer Connections (arxiv.org)

Self-Conditioning Enables Deep Sequential Reasoning in Flow Matching (arxiv.org)

Recall What Is Missing: Composing Evidence from Long-Term Agent Memory (arxiv.org)

Large Decision Models: Repurposing Large Language Models as Single-Pass Decision Makers (arxiv.org)

Data Portability Revisited: Toward the Human-Centric, AI-Driven Data Ecosystems of Tomorrow (arxiv.org)

Agent-Canary-Mas: End-to-End Evaluation of Protocol Risk in Long-Horizon Multi-Agent Systems (arxiv.org)

Self-Representation Alignment in Diffusion Models for Video Background Music Generation (arxiv.org)

Inverse Garment Design with Visual Agents (arxiv.org)

CondMSP: Condition-Informed and Resource-Aware Inorganic Materials Synthesis Planning (arxiv.org)

From Stages to States: Unified Iterative Reasoning for MLLMs (arxiv.org)

Big-Big-SycoBench: A Benchmark to Detect Sycophancy Through Model Responses to Valid vs. Invalid Arguments (arxiv.org)

Model What Matters: Decision-Guided Selective World Modeling for Language Agents (arxiv.org)

HACHIMI-II: From Role-Play to Learner Behavior Modeling for Faithful Student Simulation (arxiv.org)

MCAIR: Monitoring and Correcting Agent Unfaithfulness via Internal Representations (arxiv.org)

Learning Individual Scientific Taste with Privileged Hindsight Distillation (arxiv.org)

Non-asymptotic Convergence of Average-reward Q-learning with Options (arxiv.org)

Articulation Upper Bounds for Language Models (arxiv.org)

Discriminative Score Function: Turning Pretrained Models into Functional Generative Priors (arxiv.org)

More Methods, Better Agents? A Reality Check on Self-Distillation in Agentic Reinforcement Learning (arxiv.org)

StateCraft: Augmenting Video World Models with Abstract-State Reasoning (arxiv.org)

Average-Reward Policy Identification from a Single Continuing Trajectory (arxiv.org)

Knowledge Interaction in Hypernetwork-based Continual Learning: Conflict and Composition (arxiv.org)

FARL: A Fully Automated Pipeline of RL Interface and Solver Design (arxiv.org)

SpecForge: Recursive Self-Improvement for Coding Agents via Behavioral Specifications (arxiv.org)

Suggested Is Not Superseded: A Pre-Registered Evaluation of a System-One Model as a Memory-Retirement Gate (arxiv.org)

MedHarnessBench: A Unified Benchmark for Healthcare Agent Evaluation (arxiv.org)

History Replays Itself: Unleashing Long-Context Capability in Linear Attention (arxiv.org)

Visible but Not Avoided: State-Dependent Re-Proposal of Rejected Candidates in Catalog-Based LLM Optimization Loops (arxiv.org)

Google Earth AI Geospatial Reasoning Agent: Capabilities and a Systematic Evaluation (arxiv.org)

ReproBAIT: Can coding agents build their own biological AI tools? (arxiv.org)

JointServe: Serving Multi-Model Agents on Memory-Constrained GPUs (arxiv.org)

Into the Unknown: Benchmarking AI agents on open scientific research problems (arxiv.org)

Agent Sociodrama: Social-Dynamics-Guided Persona Recasting for LLM Agent Teams (arxiv.org)

Passing Tests, Missing Semantics: Evidence of Interpretive Lock-In in Plan-Based DataFrame Code Generation (arxiv.org)

Attention Sinks in VLMs Are Borrowed, Not Seen (arxiv.org)

Engram Injection: A Study of Temporal Validity in Parametric Language-Model Memory (arxiv.org)

RetryBench: Separating Retry Permission from Recovery Outcomes in GPT Tool Use (arxiv.org)

Comparative Study of Canvas- and Coordinate-Based Geometry-Conditioned Neural Networks for Real-Time Prediction of Indoor Environments with Varying Room Dimensions (arxiv.org)

LURE: a Phase-Labelled Synthetic Corpus of Online-Grooming Exchanges (arxiv.org)

Multi-Horizon Explainable Queue Forecasting and Resource-Aware Decision Support for Intelligent Border Control Using LightGBM (arxiv.org)

Protein Crystallization Segmentation and Classification Using Subordinate Color Channel in Fluorescence Microscopy Images (arxiv.org)

Co-DSL: Human-LLM Co-design of Reusable Spatial Concepts for Layout Generation (arxiv.org)

HGA-Quant: Heterogeneity-Aware Quantization for AutoRegressive Visual Generation Models (arxiv.org)

From Associations to Activations: Comparing Behavioral and Hidden-State Semantic Geometry in LLMs (arxiv.org)

WisdomNet: trustable machine learning toward error-free classification (arxiv.org)