Diversity is Not Ambiguity: Toward Accurate and Efficient Ambiguity Detection for Open-Domain QA

2026-08-04

AI Summary

This work addresses the limitations of existing open-domain question answering systems, which often misinterpret answer diversity as semantic ambiguity and suffer from computational inefficiency. The authors propose ARCHIVE, a novel framework that redefines ambiguity detection by leveraging logical inconsistency as the core criterion: a question is deemed ambiguous if multiple valid answers cannot coexist under a unified interpretation. ARCHIVE integrates a lightweight early-exit encoder, a conflict reasoning module, and an invariance-aware optimization objective. Evaluated on QuireQA—a newly curated benchmark comprising 4,703 queries—the framework substantially outperforms current methods, achieving F1 score improvements of up to 10.4% for ambiguity detection and 21.6% for non-ambiguity detection, while accelerating inference by 16×.

Abstract

How can question answering (QA) systems determine whether a query is ambiguous? Ambiguity detection is essential in open-domain QA, as misclassification leads to answering the wrong interpretation or unnecessary clarification. However, existing methods conflate answer diversity with ambiguity, leading to inaccurate predictions. They also process queries uniformly, resulting in wasteful computation. We propose ARCHIVE (Ambiguity Recognition via Cascaded Hypothesis Inspection and Conflict Verification), an accurate and efficient framework that detects ambiguity via logical conflict: a query is ambiguous when its valid answers cannot all be true under a single interpretation. ARCHIVE combines a lightweight early-exit encoder for surface-detectable cases with a conflict reasoning module that models logical relations among answers, reinforced by an invariance objective for robustness to noisy answer sets. We present QuireQA, a 4,703-query benchmark spanning factoid, non-factoid, and ill-formed queries. Experiments show ARCHIVE outperforms competitors, improving F1-amb by up to 10.4% and F1-unamb by up to 21.6%, while operating 16$\times$ faster than the best competitor.

Problem

• ambiguity detection
• open-domain QA
• answer diversity
• query interpretation
• computational efficiency

Innovation

• ambiguity detection
• logical conflict
• early-exit encoder
• invariance objective
• open-domain QA