Search papers, labs, and topics across Lattice.
This paper introduces the problem of feedback adaptation in RAG systems, focusing on how effectively and quickly corrective feedback propagates to future queries. They propose two evaluation metrics, correction lag and post-feedback performance, to measure this adaptation. The authors find that training-based approaches exhibit a trade-off between correction lag and reliable adaptation, and introduce PatchRAG, an inference-time method that demonstrates immediate correction and strong post-feedback generalization.
RAG systems can be patched at inference time to immediately incorporate user feedback and generalize to related queries, outperforming retraining-based approaches on adaptation speed and reliability.
Retrieval-Augmented Generation (RAG) systems are typically evaluated under static assumptions, despite being frequently corrected through user or expert feedback in deployment. Existing evaluation protocols focus on overall accuracy and fail to capture how systems adapt after feedback is introduced. We introduce feedback adaptation as a problem setting for RAG systems, which asks how effectively and how quickly corrective feedback propagates to future queries. To make this behavior measurable, we propose two evaluation axes: correction lag, which captures the delay between feedback provision and behavioral change, and post-feedback performance, which measures reliability on semantically related queries after feedback. Using these metrics, we show that training-based approaches exhibit a trade-off between delayed correction and reliable adaptation. We further propose PatchRAG, a minimal inference-time instantiation that incorporates feedback without retraining, demonstrating immediate correction and strong post-feedback generalization under the proposed evaluation. Our results highlight feedback adaptation as a previously overlooked dimension of RAG system behavior in interactive settings.
