Yoav Goldberg on new retrieval mechanism vs original RAG

yoavgo · x · 2026-08-20

Yoav Goldberg explains that his new approach differs from original RAG by training from scratch and inducing the knowledge base concurrently, offering a more expressive learned retriever than earlier LMLM attempts.

Related event: Researchers Unveil New Paradigm: Training from Scratch with Externalized Knowledge(2 posts)→

Original post →

More from Research

Research channel →