论文精读

27 篇大模型核心论文精读：Transformer、GPT-3、Scaling Law、InstructGPT、DPO、DeepSeek-R1、LoRA、FlashAttention、vLLM 等，讲清问题、方法、结论与面试考点。

• Transformer：Attention Is All You Need
• BERT：BERT: Pre-training of Deep Bidirectional Transformers
• GPT-3：Language Models are Few-Shot Learners
• Scaling Laws：Scaling Laws for Neural Language Models
• Chinchilla：Training Compute-Optimal Large Language Models
• InstructGPT：Training language models to follow instructions with human feedback
• DPO：Direct Preference Optimization
• DeepSeekMath / GRPO：DeepSeekMath: Pushing the Limits of Mathematical Reasoning
• DeepSeek-R1：DeepSeek-R1: Incentivizing Reasoning Capability via RL
• DAPO：DAPO: An Open-Source LLM Reinforcement Learning System at Scale
• Constitutional AI：Constitutional AI: Harmlessness from AI Feedback
• LoRA：LoRA: Low-Rank Adaptation of Large Language Models
• QLoRA：QLoRA: Efficient Finetuning of Quantized LLMs
• FlashAttention：FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
• vLLM / PagedAttention：Efficient Memory Management for LLM Serving with PagedAttention
• 投机解码：Fast Inference from Transformers via Speculative Decoding
• RoPE：RoFormer: Enhanced Transformer with Rotary Position Embedding
• LLaMA：LLaMA: Open and Efficient Foundation Language Models
• DeepSeek-V3：DeepSeek-V3 Technical Report
• YaRN：YaRN: Efficient Context Window Extension of LLMs
• Chain-of-Thought：Chain-of-Thought Prompting Elicits Reasoning in LLMs
• ReAct：ReAct: Synergizing Reasoning and Acting in Language Models
• RAG：Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
• Qwen3：Qwen3 Technical Report
• ViT：An Image is Worth 16x16 Words: Transformers for Image Recognition
• CLIP：Learning Transferable Visual Models From Natural Language Supervision
• LLaVA：Visual Instruction Tuning