hymie122/RAG-Survey

Organizes RAG research across three foundational integration patterns—query-based, latent representation-based, and logit-based retrieval—plus emerging speculative decoding approaches, with curated papers spanning code generation, knowledge QA, diffusion models, and multimodal synthesis. The taxonomy maps how retrieved context flows into LLM inference at different computational stages: token generation, embedding spaces, probability distributions, and decoding acceleration. Continuously updated to track rapid methodology evolution across vision, audio, and language domains beyond traditional text generation.

1,789 stars. No commits in the last 6 months.