Controllable Evidence Selection in Retrieval-Augmented Question Answering via Deterministic Utility Gating

ArXi:2603.18011v1 Announce Type: new Many modern AI question-answering systems convert text into vectors and retrieve the closest matches to a user question. While effective for topical similarity, similarity scores alone do not explain why some retrieved text can serve as evidence while other equally similar text cannot. When many candidates receive similar scores, systems may select sentences that are redundant, incomplete, or address different conditions than the question requires.