REFRAG with Xiaoqiang Lin - Weaviate Podcast #130!

Hey everyone! Thank so much for watching the 130th episode of the Weaviate Podcast featuring Xiaoqiang Lin! Xiaoqiang Lin is a Ph.D. student at the National University of Singapore. During his time at Meta, Xiaoqiang lead the research behind REFRAG: Rethinking RAG-based Decoding. Traditional RAG systems use vectors to retrieve relevant context with semantic search, but then throw away the vectors when passing the context to the LLM. REFRAG instead feeds the LLM these pre-compute vectors, achieving massive gains in long context processing and LLM inference speed! REFRAG makes Time-To-First-Token (TTFT) 31x faster and Time-To-Iterative-Token (TTIT) 3x faster, boosting overall LLM throughput by 7x while also being able to handle much longer contexts! There are so many interesting aspects to this and I really loved diving into the details with Xiaoqiang! I hope you enjoy the podcast! Links: REFRAG: Rethinking RAG-based Decoding: https://arxiv.org/pdf/2509.01092 REFRAG Explained: https://www.youtube.com/watch?v=Ek0tZootK00 Check out more works from Xiaoqiang Lin: https://xqlin98.github.io/ Chapters 0:00 Welcome Xiaoqiang! 0:43 An Introduction to REFRAG 5:59 The limits of REFRAG compression 13:07 Redundant Information in RAG 22:01 Vector Databases 27:32 Chunk Expansion Policy with RL 30:57 4 Stage Training Algorithm 38:49 How to use REFRAG LLMs? 42:01 Downstream Task Experiments 48:41 TTFT vs. Throughput Latency 52:43 MRL and REFRAG Chunk Schemas 56:27 Future Directions for AI Testing chapters fix