Blog
No more endless PDFs. Discover the core value of the latest top-tier research in one article.
Gym-Anything: Turn any Software into an Agent Environment
In-Place Test-Time Training
Skill0 : In-Context Agentic Reinforcement Learning for Skill Internalization
TriAttention: Efficient Long Reasoning with Trigonometric KV Compression
Impact of Social Media on Society
Capability Salience Vector: Fine-grained Alignment of Loss and Capabilities for Downstream Task Scaling Law
The Perils & Promises of Fact-checking with Large Language Models
Dynamic parallel tree search for efficient llm reasoning
A Law Reasoning Benchmark for LLM with Tree-Organized Structures including Factum Probandum, Evidence and Experiences
Evidential Deep Learning to Quantify Classification Uncertainty
Neural Computers
WRAP++: WEB DISCOVERY AMPLIFIED PRETRAINING
The Illusion of Superposition? A Principled Analysis of Latent Thinking in Language Models
The Art of Building Verifiers for Computer Use Agents
A Statistical X-ray Upsampling Technique for Enhanced Spatial Resolution in Computed Tomography
Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering
End-to-End Training for Unified Tokenization and Latent Denoising
Interleaved Head Attention
SkillClaw: Let Skills Evolve Collectively with Agentic Evolver
Bridging SFT and RL: Dynamic Policy Optimization for Robust Reasoning
Think Less, Know More: State-Aware Reasoning Compression with Knowledge Guidance for Efficient Reasoning
Visually-Guided Policy Optimization for Multimodal Reasoning
LatentAudit: Real-Time White-Box Faithfulness Monitoring for Retrieval-Augmented Generation with Verifiable Deployment
All elementary functions from a single operator
Nexus: Same Pretraining Loss, Better Downstream Generalization via Common Minima
R-Zero: Self-Evolving Reasoning LLM from Zero Data
ELT: Elastic Looped Transformers for Visual Generation
Visually-Guided Policy Optimization for Multimodal Reasoning
The berkeley function calling leaderboard (bfcl): From tool use to agentic evaluation of large language models
Faithfulrag: Fact-level conflict modeling for context-faithful retrieval-augmented generation