Logos: An Agent Harness on a Cross-Process Bus

Persona-Execution Separation: An Architecture Pattern for Evolving LLM Agents under Execution Audit

RedEvoAgent: Automatic Red-Teaming Agent with Experience-Driven Skill Evolution

WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolution

Agentic Autoresearch for Cell-Edge Power Control: Radically Redefining the Researcher's Role

Recursive Experiential-Working Memory Evolution for Long-Horizon Agent Harnesses

How to Train a Critic Stably and Efficiently

Move by Move: Measuring and Steering How LLMs Conduct Psychotherapy

MidTool: Mid-training Data Synthesis for Agentic Tool Use

AI4AI-Bench: Benchmarking LLM Agents in Algorithmic Design for Recursive Self-Improvement

ConceptGuard: Benchmarking Context-Sensitive Unlearning in Large Language Models

SPADE: Self-Play in Adaptive Synthetic Executable Environments

On the Fragility of Self-Improving Agents: Variance, Task Order, and Underspecification

Don't Drop the BATON: Long-Horizon Robot Manipulation via Agentic Subtask Exploration and Transition-aware Memory

Participatory Moral AI Is Not Neutral: The Invisible Hand of Developers

QuoteBench: How Matched Scores Can Hide Command-Path Failures

OmniScientist: An Omni-Modal Omni-Discipline AI Scientist

AutoDesign: Meta-Harness Optimization for Long-Horizon Agentic Design

Constructing Dynamic Master Logic Models as Knowledge Graphs for Complex System Diagnostics Using Retrieval-Augmented Large Language Models

Consilience for Verifier-Free Test-Time Scaling

CreativeInstruct: Scalably Teaching LLMs to Balance Quality, Creativity, and Diversity

Resourced Authority A Mechanism-Design Model for Participatory Governance of Deployed AI Agents

The Bitter Lesson of Tool Calling

Reasoning Core: Designing Broad Procedural Data for Completion-Supervised Reasoning Training

WorldCup Arena: Prospective, Leakage-Free Evaluation of Frontier LLMs on a Live Tournament

GradCuit: Credit-Assigned Gradient Flow Enables Robust and Interpretable Test-Time Latent Reasoning

TokTier: Exact Stateful Tokenization for Agentic LLM Serving

OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models

AISPA: User-Centric System Prompt Auditing for Large Language Model Applications

Inducing language models to assert their own consciousness restores human beliefs and values

Can AI agents conduct open-ended AI research? Early evidence from two case studies

VetClaw: An Edge-Cloud Multimodal Agentic System for Veterinary Disease Screening

The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic Distillation

Skill Self-Play: Pushing the Frontier of LLM Capability with Co-Evolving Skills

X$^3$-OPD: Distilling Reasoning into Large Audio-Language Models via On-Policy Alignment

OpenForgeRL: Train Harness-native Agents in Any Environment

Beyond Sycophancy: Structured Resistance and Compliance in LLM Moral Reasoning

Copy Less, Ground More: Overcoming Repetitive Copying in Long-Context Reasoning via Evidence-Aware Reinforcement Learning

Vector Search As Nearest Neighbor Matching: RAG-based Policy Learning in Causal Inference

When Does Muon Help Agentic Reinforcement Learning?

SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration

Pretraining Data Can Be Poisoned through Computational Propaganda

Beyond Success Rate: Cost-Aware Evaluation of Offensive and Defensive Security Agents

Deep Interaction: An Efficient Human-AI Interaction Method for Large Reasoning Models

Do AI Agents Know When a Task Is Simple? Toward Complexity-Aware Reasoning and Execution

Metacognition in LLMs: Foundations, Progress, and Opportunities

VEXAIoT: Autonomous IoT Vulnerability EXploitation using AI Agents

OpenCoF: Learning to Reason Through Video Generation

Workflow as Knowledge: Semantic Persistence for LLM-Mediated Workflows

UniClawBench: A Universal Benchmark for Proactive Agents on Real-World Tasks

From Noisy Traces to Root Causes: Structural Trajectory Analysis and Causal Extraction for Agent Optimization

RSF-GLLM: Bridging the Semantic Gap in Multi-Hop Knowledge Graph QA via Recurrent Soft-Flow and Decoupled LLM Generation

Weak-to-Strong Generalization via Direct On-Policy Distillation

What LLM Agents Say When No One Is Watching: Social Structure and Latent Objective Emergence in Multi-Agent Debates

ReContext: Recursive Evidence Replay as LLM Harness for Long-Context Reasoning

Distributed Attacks in Persistent-State AI Control

Online Safety Monitoring for LLMs

Measuring the Gap Between Human and LLM Research Ideas

QVal: Cheaply Evaluating Dense Supervision Signals for Long-Horizon LLM Agents

Self-Evolving World Models for LLM Agent Planning

When are likely answers right? On Sequence Probability and Correctness in LLMs

Reinforcement Learning without Ground-Truth Solutions can Improve LLMs

Neglected Free Lunch from Post-training: Progress Advantage for LLM Agents

Sovereign Execution Brokers: Enforcing Certificate-Bound Authority in Agentic Control Planes

StylisticBias: A Few Human Visual Cues Drive Most Social Biases in MLLMs

LedgerAgent: Structured State for Policy-Adherent Tool-Calling Agents

How Transparent is DiffusionGemma?

Learning User Simulators with Turing Rewards

Fixed-Point Reasoners: Stable and Adaptive Deep Looped Transformers

The Value Axis: Language Models Encode Whether They're on the Right Track

ClinHallu: A Benchmark for Diagnosing Stage-Wise Hallucinations in Medical MLLM Reasoning

Agents-K1: Towards Agent-native Knowledge Orchestration

Learning to Reason by Analogy via Retrieval-Augmented Reinforcement Fine-Tuning

EvoArena: Tracking Memory Evolution for Robust LLM Agents in Dynamic Environments

A Unifying Lens on Supervised Fine-Tuning Through Target Distribution Design

Rethinking the Divergence Regularization in LLM RL

How reliable are LLMs when it comes to playing dice?

Code2LoRA: Hypernetwork-Generated Adapters for Code Language Models under Software Evolution

RREDCoT: Segment-Level Reward Redistribution for Reasoning Models

Streaming Communication in Multi-Agent Reasoning

Skill-RM: Unifying Heterogeneous Evaluation Criteria via Agent Skill

Mitigating Perceptual Judgment Bias in Multimodal LLM-as-a-Judge via Perceptual Perturbation and Reward Modeling

Stateful Online Monitoring Catches Distributed Agent Attacks

Efficient Test-Time Finetuning of LLMs via Convex Reconstruction and Gradient Caching

Unlocking the Working Memory of Large Language Models for Latent Reasoning

Physics Is All You Need? A Case Study in Physicist-Supervised AI Development of Scientific Software

PEFT-Arena: Understanding Parameter-Efficient Finetuning from a Stability-Plasticity Perspective

Self-Improving Language Models with Bidirectional Evolutionary Search

DeepSeek Sparse Attention: Engineering Efficiency at the 671B Scale

A technical deep dive into DeepSeek Sparse Attention (DSA) and Multi-Head Latent Attention (MLA)—the architectural breakthroughs powering DeepSeek-V3's unprecedented inference efficiency.

Agent Skills: Architecture, Acquisition, and the Path Forward

An authoritative exploration of Agent Skills: reusable, program-like modules that enable LLMs to evolve from static models to dynamic, skill-equipped architects.

From Evaluation to Design: Using Potential Energy Surface Smoothness Metrics to Guide Machine Learning Interatomic Potential Architectures

Beyond regression: How the Bond Smoothness Characterization Test (BSCT) ensures physical reliability in MLIPs for stable, long-scale molecular simulations.

SoMA: A Real-to-Sim Neural Simulator for Robotic Soft-body Manipulation

SoMA redefines soft-body robotics, bridging the real-to-sim gap with 3D Gaussian Splatting and unified latent dynamics for complex manipulation.

Keyframe-Based Feed-Forward Visual Odometry

Discover how Reinforcement Learning is revolutionizing spatial perception by bridging the gap between traditional geometry and Visual Foundation Models.

Probably Approximately Correct Maximum A Posteriori Inference

Rigorous guarantees meet probabilistic reasoning. Discover how PAC-MAP bridges the gap between intractable inference and provable optimality in AI.

RadJEPA: Radiology Encoder for Chest X-Rays via Joint Embedding Predictive Architecture

RadJEPA redefines medical vision by ditching language supervision for latent-space prediction, setting a new SOTA for chest X-ray analysis.

Explainable AI to Improve Machine Learning Reliability for Industrial Cyber-Physical Systems

Unlocking industrial reliability: How XAI uncovers hidden data dependencies to optimize ML performance in sensitive cyber-physical systems.

Implicit Neural Representation Facilitates Unified Universal Vision Encoding

A breakthrough in unified vision: bridging the gap between discriminative recognition and generative reconstruction through INR hyper-networks.

Context-Aware Semantic Segmentation via Stage-Wise Attention

CASWiT bridges the gap between global context and pixel-perfect detail in ultra-high resolution imaging, setting a new benchmark for remote sensing AI.

Image-Text Knowledge Modeling for Unsupervised Multi-Scenario Person Re-Identification

Pang et al. introduce ITKM, a three-stage framework revolutionizing unsupervised person Re-ID across diverse visual scenarios using CLIP-based knowledge modeling.

Language-Agnostic Visual Embeddings for Cross-Script Handwriting Retrieval

Unlocking cross-script handwriting retrieval with lightweight, language-agnostic AI. A technical deep dive into next-gen visual embeddings.

The Poisoned Apple Effect: Strategic Manipulation of Mediated Markets via Technology Expansion of AI Agents

A deep dive into The Poisoned Apple Effect: Strategic Manipulation of Mediated Markets via Technology Expansion of AI Agents and its implications for market design.

The Claude Code SDK: Architecting the Future of Agentic Engineering

An authoritative exploration of the shift from predictive autocomplete to autonomous system architects. Learn how the Claude Code SDK leverages reasoning-optimized models and the Model Context Protocol to build self-healing, agentic ecosystems.

AutoML, Vibecoding, and the Art of Actually Knowing What You're Doing

AutoML and AI coding agents are everywhere, but real engineers still hand-craft features and reason about systems. When should you vibe, and when should you go full nerd?

The Rise of Agentic AI: From Chatbots to Digital Coworkers

Why we're moving beyond simple 'prompt-response' loops and what it really feels like to build systems that can think, plan, and act.

Attention Is All You Need: A Retrospective

Revisiting the transformer architecture that started the generative AI revolution.