Notes on LLMs, deep learning, and building things that work.

2026

• CLIP: Learning Transferable Visual Models From Language Supervision 07/11

• Self-Consistency: Improving Chain-of-Thought Reasoning 07/10

• Constitutional AI: Harmlessness from AI Feedback 07/09

• DPO: Direct Preference Optimization 07/07

• RLHF: Deep Reinforcement Learning from Human Preferences 07/05

• RoPE: Rotary Position Embedding 07/03

• Chinchilla: Training Compute-Optimal Large Language Models 06/30

• Scaling Laws for Neural Language Models 06/28

• Tree of Thoughts: Deliberate Problem Solving 06/21

• AutoGen: Multi-Agent Conversation as Computation 06/14

• Generative Agents: Believable Simulacra of Human Behavior 06/07

• Reflexion: Verbal Reinforcement Learning for Language Agents 05/31

• Toolformer: Language Models That Teach Themselves to Use Tools 05/24

• ReAct: Reasoning + Acting in Language Models 05/17

• RAG: Retrieval-Augmented Generation 05/10

• DeepSeek-R1: Reasoning via Reinforcement Learning 05/03

• Mixtral of Experts: Sparse Mixture of Experts 04/26

• LLaMA: The Open-Weight Catalyst 04/19

• FlashAttention: IO-Aware Exact Attention 04/12

• LoRA: Low-Rank Adaptation of Large Models 04/05

• Chain-of-Thought Prompting: Let the Model Think Step by Step 03/29

• InstructGPT: Aligning LLMs with Human Preference (RLHF) 03/22

• GPT-3 and Few-Shot Learning at Scale 03/15

• BERT: Bidirectional Pretraining from Transformers 03/08

• The Transformer: Attention Is All You Need 03/01

• Deploy OpenClaw AI on Raspberry Pi: Our Open Source Guide 02/04

• Homeware Sense: Open Source Smart Home Integration for OpenClaw AI 02/02

• wechat-dev-mcp: Bridging AI and WeChat Mini-Program Development 01/31

• Computer Vision 01/15

• Advanced SQL 01/10

• Geospatial Analysis 01/02

2025

• Machine Learning Explainability 12/28

• Game AI and Reinforcement Learning 12/20

• Deep Learning 10/22

• Kaggle与数据科学：实战技巧与最佳实践 09/15