Gap Discussion
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
What did this paper leave unanswered? Identify research gaps, propose solutions, and connect with others building implementations.
Identify a Research Gap
0/2000 characters
1 Discussion
JK
Jordan Kim
62
The original RAG paper evaluates on open-domain QA benchmarks. In production, the hard problem is retrieval quality at scale. With 10M+ documents, recall@5 drops significantly. The paper doesn't address: 1) Multi-vector representations (ColBERT), 2) Hierarchical indexing, 3) Query expansion. What's the state of the art for high-recall retrieval over large corpora?