rag-all-techniques and RAG-Arena
These are complements: one provides simplified implementations of multiple RAG techniques for practical application, while the other offers comparative evaluation and explanation of those same techniques, making them useful together for both learning and benchmarking RAG approaches.
About rag-all-techniques
liu673/rag-all-techniques
Implementation of all RAG techniques in a simpler way（以简单的方式实现所有 RAG 技术）
Implements 17+ RAG techniques (semantic chunking, query transformation, reranking, graph-based retrieval, etc.) using standard Python libraries (OpenAI, NumPy, PyMuPDF) rather than framework abstractions like LangChain or FAISS. Each technique includes fully-commented notebook implementations demonstrating the complete pipeline from document ingestion through embedding creation, semantic search, and LLM-based response generation. Covers advanced patterns including adaptive retrieval strategy selection, self-evaluating RAG with relevance assessment, hybrid vector/BM25 fusion, and iterative feedback loops for continuous optimization.
About RAG-Arena
ZehaoJia1024/RAG-Arena
讲解并评估多种RAG算法
Scores updated daily from GitHub, PyPI, and npm data. How scores work