Paper Reviews
Deep teardowns of key AI/ML papers — LLMs, multimodal, fine-tuning, RAG — down to their design decisions and trade-offs, grouped by topic.
Language Models
- GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning - November 17, 2026
- Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs - November 17, 2026
- DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines - November 17, 2026
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models - January 12, 2026
- Qwen3 Technical Report - October 8, 2025
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning - October 7, 2025
- DeepSeek-V3 Technical Report - October 7, 2025
- Gemma 2: Improving Open Language Models at a Practical Size - October 7, 2025
- The Llama 3 Herd of Models - October 7, 2025
- Tulu 3: Pushing Frontiers in Open Language Model Post-Training - October 6, 2025
- DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model - October 5, 2025
- Kimi K2: Open Agentic Intelligence - October 3, 2025
- Mixtral of Experts - April 4, 2025
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models - February 28, 2025
- Mistral 7B - February 27, 2025
- DeepSeek-Coder: When the Large Language Model Meets Programming - The Rise of Code Intelligence - February 14, 2025
- DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models - February 5, 2025
- DeepSeek LLM: Scaling Open-Source Language Models with Longtermism - February 3, 2025
- LLaMA: Open and Efficient Foundation Language Models - January 28, 2025
- Llama 2: Open Foundation and Fine-Tuned Chat Models - January 27, 2025
Multimodal Learning
- Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities - October 11, 2025
- Gemini: A Family of Highly Capable Multimodal Models - October 11, 2025
- Pixtral 12B - October 10, 2025
- Gemma 3 Technical Report - October 3, 2025
Finetuning
- Constitutional AI: Harmlessness from AI Feedback - January 16, 2026
- QLoRA: Efficient Finetuning of Quantized LLMs - October 24, 2025
- Direct Preference Optimization: Your Language Model is Secretly a Reward Model - October 19, 2025
Retrieval Augmented Generation
- Zep: A Temporal Knowledge Graph Architecture for Agent Memory - June 7, 2026
- LightRAG: Simple and Fast Retrieval-Augmented Generation - March 25, 2026
- RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval - February 14, 2025
- From Local to Global: A Graph RAG Approach to Query-Focused Summarization - January 17, 2025