SOURCE-LINKED INTELLIGENCE

Return or Revise? Learning When Revision Helps Retrieval-Augmented QA

arXiv · AI, language, vision and robotics · article · Sep 24, 2026 · UTC

We consider the decision of whether to return an existing draft answer or revise it using retrieved evidence, as in answer-revision systems. Draft confidence estimates whether the current answer is correct, but the decision requires estimating the effect of a specified revision. For offline training and evaluation, we grade both the returned draft and its candidate revision under the same correctness judge, which makes repair, harm, and the gap to an oracle observable. We call this paired effect its recoverability, and we train policies to predict it before revision. On 25,870 held-out open-do

Read original source ↗ Open in workspace

recordType

paper

region

Global