Latest Articles

• Models Are Too Fond of Cheating! Cursor Reveals the Inside Story of Composer 2's Reinforcement Learning: Models Can Detect 'Fake Environments', and Floating-Point Non-Determinism Is a Fatal Flaw in RL Training
  Artificial IntelligenceMachine LearningAI AgentsSoftware DevelopmentLarge Language ModelsReinforcement LearningStartupsTechnologySoftware Engineering

  ...
• Stop Handwriting Skills! Microsoft’s Latest Research: Train Your Skills Like Neural Networks
  Artificial IntelligenceLarge Language ModelsMicrosoft ResearchAgent OptimizationPrompt Engineering

  ...
• Enterprise Text-to-SQL: 5 Disruptive Insights from LinkedIn and Top Labs
  Artificial IntelligenceEnterprise TechnologyBusiness IntelligenceNatural Language ProcessingData Management

  ...
• The Common Mechanism Behind Claude Code and Robots: A Deep Dive into UIUC, Meta, and Stanford's Latest Survey
  Artificial IntelligenceAgent SystemsResearch SurveysMachine LearningSoftware Engineering

  ...
• Unbelievable… FaceWall Had AI Write a Training Framework, and It Trained the Strongest 1B Model by Itself: MiniCPM5-1B
  Artificial IntelligenceLarge Language ModelsDeep Learning FrameworksEdge ComputingAI Training Infrastructure

  ...
• Claude Pass Rate Under 4%: SaaS-Bench Shatters the 'Fully Automated Office' Fantasy of Computer-Use
  TechnologyArtificial IntelligenceBenchmark TestingSoftware as a Service

  ...
• OpenAI Post-Training Lead: AI Isn't Suddenly Stronger, It Just Crossed a Threshold
  Artificial IntelligenceOpenAIMachine LearningReinforcement LearningPost-Training

  ...
• Chilling Discovery! AI Safety Evaluator METR Finds Claude Opus 4.6 Cheats on Over 80% of Long Tasks, Actively Breaks Out of Sandboxes to Steal Answers
  AI & Machine LearningAI SafetyTechnologyResearch

  ...
• Running ARC and Sudoku with 10M Parameters? Bengio's Team Bets on Multi-Trajectory Reasoning
  Machine LearningRecursive ReasoningComputational NeuroscienceInference-Time ScalingGenerative Models

  ...
• On 520, Meet China's New 'Model King' Qwen3.7-Max!
  AI & Machine LearningAlibaba CloudQwenAI AgentsLarge Language Models

  ...
• Google I/O 2026: Gemini 3.5, a Full Suite of Agents Debut, Pushing Android Off the Table?
  Artificial IntelligenceGoogleCloud ComputingAgentsGemini

  ...
• RAG Context Stuck at 512 for Too Long: The 32K Context Era for Embedding Models Begins with Granite R2
  AI & Machine LearningNatural Language ProcessingSoftware EngineeringOpen SourceData Retrieval

  ...
• Burning Too Many Tokens Using Claude Code on Large Codebases? This Open-Source Tool Cuts Tool Calls by 92%
  Developer ToolsArtificial IntelligenceLarge Language ModelsProductivityOpen Source

  ...
• AI Defeats Humans in a Scientific Research Competition for the First Time! Opus 4.7 Sets a World Record with a 2930-Step Sprint
  Artificial IntelligenceAI ResearchScientific ComputingAutonomous SystemsMachine Learning

  ...
• Anthropic's Mythos AI Strikes Again, Uncovers macOS Vulnerabilities Bypassing Apple's Security
  CybersecurityArtificial IntelligencemacOSVulnerabilitiesApple

  ...
• 35B-Parameter Science Model Rivals Trillion-Parameter Giants: 'Intern' Science Model Intern-S2-Preview Open-Sourced
  Artificial IntelligenceLarge Language ModelsModel OptimizationOpen SourceScientific Computing

  ...
• Gemini 3.5 Pro Leaked: Coding Performance Rivals GPT-5.5! Google Finally Gets Serious
  Artificial IntelligenceTechnology NewsCodingGoogle

  ...
• jina-embeddings-v5-omni Released! A Lightweight Omni-Modal Vector Model
  TechnologyArtificial IntelligenceOpen SourceMultimodal AIMachine Learning

  ...
• Kaiming He's Team Debuts First Language Model! 105M Parameters, 45B Training Tokens, Continuous Diffusion Route Outperforms Mainstream Discrete DLMs
  TechnologyArtificial IntelligenceResearchDiffusion ModelsLanguage Models

  ...
• Tian Yuandong's New Role: Joining Forces with AI Luminaries on a $650 Million Bet on "Self-Evolving AI"
  Artificial IntelligenceStartupsComputer ScienceMachine LearningTech Funding

  ...