Category: Machine Learning
- 500 Seed Samples, Four Self-Evolving Agents: Reasoning Capability Surges by 10.7%
- Reconstructing Native Multimodality! Meituan Releases Purely Discrete Base Model, Truly Achieving 'Everything is Token'
- Models Have Gained Introspective Capabilities, But Their Inner Doors Were Locked | Hao's Paper Talk
- Anthropic Engineering Blog: How Anthropic Designed Claude Code's Auto Mode
- Stunning Reversal in World's T hardest Exam: Dark Horse AI Breaks 36% Barrier as Top Models Crash
- World's First AI Scientist Publishes in Nature: Mastering the Entire Research Process from Idea to Paper, Passing Blind Human Review
- Mozilla Unveils CQ Project: A 'Stack Overflow for AI Agents'
- VideoSeek Long-Video Understanding Agent: The Secret to Boosting GPT-5's Long-Video Comprehension by 10 Points
- Anthropic Adopts GAN-Inspired Approach to Solve AI Output Quality Issues
- Overnight, AI Gains 'Permanent Memory'! Smashes SOTA with 99% on Toughest Exam, Netizens Go Wild
- NVIDIA Nemotron-Cascade 2 Technical Report
- How Can a Model Trained on 200M Real Tokens Match the Performance of 360M Data?
- Terence Tao Uses Claude Code to Solve Problems, Crashes Twice Due to Running Out of Tokens
- Latest! Karpathy's 10,000-Word In-Depth Interview: My Anxiety Became an AI Addiction—All Verifiable Domains Will Eventually Belong to Machines
- OpenAI is throwing everything into building a fully automated researcher
- Golden Rules for Skill Development! Google Releases 5 Agent Skill Design Patterns
- Performance Surges 42%! Renmin University & ByteDance Open-Source Scale-SWE, a 100k-Level SWE Dataset
- Mamba-3
- Someone Actually Built the AI Research Community from Karpathy's Side Project...
- OpenClaw-RL: Allowing AI Agents to Self-Evolve Through Chat
- a16z: Agents Perform Poorly Due to Lack of Correct Data Context
- 4B Model Surpasses GPT-5 in Hallucination Suppression: CMU and Others Propose New Behaviorally Calibrated Reinforcement Learning Method
- Jensen Huang Enters the OpenClaw Arena! Most Powerful Open-Source 'Lobster' Model Rivals Opus 4.6
- Masterpiece! MIT and Google Train an LLM Capable of Rigorous Bayesian Inference
- Karpathy Slept, AI Ran 100 Experiments for Him