ConfRAG: Confidence-Guided Retrieval-Augmenting Generation
This work introduces ConfQA, a fine-tuning strategy that reduces hallucination rates from 20-40% to below 5% across multiple factuality benchmarks, and proposes ConfRAG, a triggering strategy that invokes RAG only when the model responses with unsure.
Yin Huang, Y. Xu, Kai Sun et al.
· 2 citations