Category: Large Language Models

• Latest Interview with Jeff Dean, the Soul of Gemini and Legendary Engineer: In the Future, Everyone Will Have 50 Virtual Interns, No Need for Experts Anymore!
• Qwen3.5: Towards Native Multimodal Agents
• The Bitter Lesson! ROLL Team Shares: Practical Experience in Agentic RL Training
• Xiaomi Introduces JudgeRLVR: Judge First, Generate Second — Breaking the Efficiency Paradox of "Long Chain-of-Thought" in Reasoning Models
• ModelBest's SALA Architecture Is Tearing Down the Transformer's Wall
• Nvidia's new technique cuts LLM reasoning costs by 8x without losing accuracy
• Stable-DiffCoder Surpasses Autoregressive Models! New Breakthrough in Code Generation with Diffusion Models
• MianBi Intelligence 9B On-Device Full-Modal Open Source: See, Listen, Interrupt Anytime, Instant Interaction
• Just Released: Claude 4.6 and GPT-5.3-Codex Simultaneously Launched!
• Zhipu's New Model Also Uses DeepSeek's MLA, Runs on Apple M5
• True Hack! MIT New Research: Zero Architecture Changes, Unlocking Million-Level Context for Large Models
• Is Agentic RAG Worth It? A Four-Dimensional Real-World Test Reveals the Answer!
• From 'LLM-as-a-Judge' to 'Agent-as-a-Judge': A Review of the Three-Stage Evolution of AI Evaluation Paradigms
• Just Now, Liang Wenfeng Signs New Paper That Explodes Late at Night! DeepSeek-V4 New Architecture Revealed: Proposes New Sparsity Direction, Complements MoE, Significantly Extends Long Context Capabilities, Stronger Reasoning and Coding Abilities
• LAMER: Meta-Reinforcement Learning Enables Language Agents to Perform Active Exploration
• Stanford's Latest Course Release: Hand-Writing Code is Banned, AI is Mandatory
• From 'Titans+MIRAS & Nested' Architectural Innovations to NeurIPS2025 Best Paper 'Gated Attention'
• Causal Inference Charges into the LLM Battlefield! Large Model Hallucination Terminator? ABCA Framework
• Paper Brief | Activating and Enhancing Causal Reasoning Capabilities of Large Language Models Using Conditional Statements (CL2025)
• What? RLVR Isn't Learning New Knowledge—It's Learning How to Use Knowledge for Reasoning!
• Karpathy Forms a Large Model 'Parliament': GPT-5.1, Gemini 3 Pro, and Others Become the Strongest Think Tank
• Under $8,000! Sina Weibo's 1.5B Small Model Surpasses Near-Trillion Parameter Models
• Microsoft Proposes GAD Framework: Open-Source Models Can Directly Distill Black-Box GPT-5
• Making LLMs Work Like a Company: Microsoft Turns 'Concurrent Thinking' into a Protocol, Higher Accuracy and 28% Reduction in Critical Path Latency
• Reinforcement Learning + Large Model Memory: Mem-α, Enabling Agents to "Learn How to Remember" for the First Time