AI News

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

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

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

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

MemeLogos: Knowledge-Driven Retrieval-Augmented Reasoning for Harmful Meme Detection (arxiv.org)

Scalable Robust MCTS with Anytime Deterministic Optimality Guarantees (arxiv.org)

Who's Behind the Text? Robust Model Authorship Attribution (arxiv.org)

Bridging the Topological Divide: Dual-Branch Graph Distillation with Semantic Priors for Robust Cold-Start Recommendation (arxiv.org)

Geometric Kernel Diversity: An Axiomatic Diversity Measure in Euclidean Space (arxiv.org)

INTACT: Inducing Latent Reasoning on a Frozen Backbone (arxiv.org)

Hidden in Plain Sight: Advancing Harmful Content Understanding via Perception-Aware Adversarial Self-Play (arxiv.org)

SpecFit: Not All Draft Tokens Cost the Same in Offloaded MoE Inference (arxiv.org)

CoverSynth: Coverage-Guided Generation of Retrieval Evaluation Sets (arxiv.org)

State Representation Shapes Long-Horizon Agent Learning and Transfer: An Empirical Study (arxiv.org)

Evidence-Anchored Overdose Death Prediction from Death Investigation Reports (arxiv.org)

Analyzing the Dynamics of WSBM-based Echo State Networks using InputDSA (arxiv.org)

Fewer Bits, More Tries: Recovering Capability Under Memory Constraints (arxiv.org)

TRACE: Temporal Risk Assessment and Concept Editing for Safer Language Generation (arxiv.org)

MultiFlow: Vision-Driven Multimodal fMRI Encoding via Stimulus-Informed Source Flow Matching (arxiv.org)

A Unified Response Framework for Questions and Tutoring Dialogues (arxiv.org)

Digital Self-Interference Cancellation for In-Band Full-Duplex Systems Based on Transformer Neural Operators (arxiv.org)

CHEAR: Confusion-Hierarchy Evidence-Aware Reasoning for Cost-Efficient Human Activity Recognition (arxiv.org)

Learning to Select Features with Manifold Deep Unfolding PCA (arxiv.org)

Scaling LLM Inference via Parallelism: Throughput-Latency Trade-off (arxiv.org)

Hypothesis Selection via Weighted Medians (arxiv.org)

ActMOF: Mapping Reticular Synthesis through Active Learning (arxiv.org)

The Impact of Detector Architecture, Model Scale, and Scene Density on Multi-Pedestrian Tracking (arxiv.org)

Tail Competition Explains Scaling in Best-of-N Sampling for Verifiable Problems (arxiv.org)

Steering Vectors for Competing Ethical Frameworks Are Interchangeable: A Competing-Target Test for Activation Steering (arxiv.org)

Target-Token Asymmetry Can Masquerade As Depth Specialization (arxiv.org)

TopicMerge: A Divide and Conquer Approach to Topic Modeling (arxiv.org)

From Detection To Anticipation: Continuous Onset Forecasting For Stuttered Speech (arxiv.org)

Learning in Strategic Queuing with Geometric Recharge: From Bandits to Systems (arxiv.org)

ProgAgent: Learning Tool Orchestration through Programmatic Training (arxiv.org)

OntoAR-HAR: Ontology-Guided Agentic RAG with Personalized Memory for Human Activity Recognition (arxiv.org)

RECAP: Calibrating Low-Bit Reasoning Models on the History They Read (arxiv.org)

ClinEdge: An Expert-Graded Benchmark of Real Patient Edge Cases Where Frontier Multimodal Models Fail (arxiv.org)

MoMaTune: Model, Match, Then Tune Language Models for Numerical Records (arxiv.org)

Using Evidence-Centered Design to Guide Prompt Engineering in Measurement: An Illustration with Automated Essay Scoring (arxiv.org)

Typed Decision Models on Tabular Data (arxiv.org)

Words to World: Benchmarking LLMs in Interactive Physical Environments (arxiv.org)

Understanding Unified VLMs: From Attention Probing to Guidance Tuning for Multimodal Generation (arxiv.org)

HarnessNet: Feedback-driven Hierarchical Joint Harness Evolution for Multi-agent Systems (arxiv.org)

Running Fewer Loops Removes Refusal Behaviour in Recurrent-Depth Language Models (arxiv.org)

Can agents whose sole duty is to keep humans alive be hijacked to kill? Delayed multi-turn derived-action trap attacks in simulated high-stakes environments (arxiv.org)

Harmony Harness: Multi-Objective Self-Evolution of Agent Harnesses (arxiv.org)

Single-step Retrosynthesis via Reinforcement Learning from Forward Feedback (arxiv.org)

Trust Hinders Bidirectional Human-AI Alignment (arxiv.org)

GesBench: When and Where Do Multimodal Language Models Perceive Social Meaning? (arxiv.org)

Retrieval Over Training: Similarity search-based Model Selection for Time Series Anomaly Detection (arxiv.org)