Mateusz Stachowicz’s Post

A trick in retrieval-augmented generation (RAG) is to use the output of RAG as extra context information to perform second RAG query to get better retrieval result and eventually get a better output. In my interpretation it is a resampling of documents in the vector storage to get more relevant documents as the result of retrieval. And I think we can do the same by querying the vector storage using the embedding output of ReRanker for better retrieval without large cost overhead using text generation. Thanks Paul Tsoi for the paper https://lnkd.in/g8NFdZBg

Pretty excited about this new RAG technique 🧑🍳 A top issue with RAG chunking is it splits the document into fragmented pieces, causing top-k retrieval to return partial context. Also most documents have multiple hierarchies of sections: top-level sections, sub-sections, etc. This is also why lots of people are interested in exploring the idea of knowledge graphs - pulling in "links" to related pages to expand retrieved context. This notebook lets you retrieve contiguous chunks without having to spend a lot of time tuning the chunking algorithm, thanks to GraphRAG-esque metadata tagging + retrieval. Tag chunks with sections, and use the section ID to expand the retrieved set. #RAGTechnique #GraphRAG #KnowledgeGraphs #AIContextualUnderstanding #InformationRetrieval #NaturalLanguageProcessing #ChunkingOptimization #ArtificialIntelligenceInnovation #MachineLearningAdvancements #LanguageModelingSolutions https://lnkd.in/gqTnfKWG

I just published Advanced RAG Retrieval Strategies: Flow and Modular

Really useful blog from Zain Hasan - "Multimodal Document RAG with Llama 3.2 Vision and ColQwen2" - worth checking out here -> https://lnkd.in/dJj6W-3T link of the notebook -> https://lnkd.in/d3UcuyBV

Graph RAG Works Better Than Standard RAG #GraphRAG leverages structural information across entities to enable more precise and comprehensive retrieval, capturing relational knowledge and facilitating more accurate, context-aware responses. This improves the accuracy of standard RAG systems.

This weekend, learn about 5 different ways of evaluating your RAG systems. zhaozhiming takes you through a comprehensive tour of the different RAG evaluation methods using LLM-as-a-judge (which have corresponding LlamaIndex implementations): 1. Answer Relevance 2. Context Relevance 3. Faithfulness 4. Correctness 5. Pairwise Comparison Along with synthetic dataset generation. Check it out here: https://lnkd.in/gdQ3fZ2W

🍍Complex RAG Flow with LangGraph LangGraph for building an advanced RAG flow using ideas from 3 papers: - Corrective-RAG (CRAG) - Self-RAG - Adaptive RAG YouTube: https://lnkd.in/gHUpiZS5

Very useful when considering RAG evaluation.