Category: Natural Language Processing

• Reconstructing Native Multimodality! Meituan Releases Purely Discrete Base Model, Truly Achieving 'Everything is Token'
• How Can a Model Trained on 200M Real Tokens Match the Performance of 360M Data?
• One Year into RAG: My Biggest Regret Was Adopting Knowledge Graphs
• Masterpiece! MIT and Google Train an LLM Capable of Rigorous Bayesian Inference
• MMLU is Dead? 'Humanity's Last Exam' Published in Nature: Global AI Models Collectively Fail!
• Google's New Research Identifies the Crucial Tokens Where Large Models Ponder Deeply!
• Is PPO Dead? The Reinforcement Learning Foundation Used by DeepSeek Has Major Flaws!
• Transformer Authors Lead Sakana AI to Release Three Papers: Completely Reconstructing Long-Text Memory Mechanisms
• Google's New Discovery: DeepSeek Reasoning Splits into Multiple Personalities, Left and Right Brain Competing for Intelligence
• True Hack! MIT New Research: Zero Architecture Changes, Unlocking Million-Level Context for Large Models
• Google Just Overturned Model Memory, and Nvidia Revolutionized Attention｜Hao Good Chat Paper
• Attention Is Not What You Need? Reframing Sequence Modeling with Geometric Aesthetics via Grassmann Manifolds
• LLMs in Document Intelligence: Survey, Progress, and Future Trends
• Meta Introduces Deep Think with Confidence: Boosting Reasoning Accuracy and Efficiency with Minimal Changes
• DeepSeek R2's Secret Weapon Revealed! The Technology Just Awarded a Top Prize to Liang Wen-feng Allows AI to Read Long Texts 11 Times Faster
• ACL 2025 | Large Models "Spreading Misinformation"? DRAG's Two-Stage "Multi-Agent Debate" Solves Hallucination on Hallucination
• Global Programmers Explode! Jensen Huang Declares in London: The Future of Programming Languages is "Human"
• Mianbi MiniCPM4: 3x Inference Speed, Outperforming Same-Size Qwen3, Putting Pressure on Alibaba
• After ZeroSearch, Tongyi's Latest Work MaskSearch Proposes a New Framework for Reasoning-Search Pre-training
• Reviewing the Progress of RL-Reasoning
• AI Learns Reasoning Solely by "Confidence": Zhejiang University Alumnus Replicates DeepSeek's Long Chain-of-Thought Emergence, Reinforcement Learning Needs No External Reward Signals
• Tsinghua University's New RAG Framework: DO-RAG Accuracy Soars by 33%!
• Qwen Team Releases Long-Context Reasoning Model QwenLong-L1, Surpassing o3-mini
• Alibaba Open-Sources New Qwen Model: A Dragon Boat Festival Gift!
• ICML 2025 | Training-Free, Instant Alignment of Large Model Preferences