新指标ConflictScore能量化大模型对矛盾证据的识别能力。
ConflictScore: Identifying and Measuring How Language Models Handle Conflicting Evidence
- 将回答拆成原子命题，逐条比对文档中的支持与矛盾证据。
- 提出两个指标：有矛盾的命题占比（CS-C）和正反证据平衡度（CS-R）。
- 适用于检测过度自信结论，适合提升问答系统真实性。
现有事实性与忠实性评估指标无法捕捉支持与反对证据共存的情况。我们提出冲突评分（ConflictScore），用于量化模型在回应中承认矛盾证据的能力。该框架将回答分解为原子主张，逐一标注其与每篇参考文档的关系，并聚合为两个互补指标：冲突评分-数量（CS-C），即存在矛盾的主张比例；冲突评分-比率（CS-R），反映支持与反对证据的平衡程度。我们构建了ConflictBench基准，涵盖模糊性、矛盾和观点分歧等多种冲突形式，以系统评估该指标。实验表明，ConflictScore能有效识别跨领域的过度自信陈述，并可作为校正反馈机制，提升TruthfulQA上的真实性。
原文摘要 · Abstract (English)
Existing metrics for factuality and faithfulness evaluate whether an answer is supported or contradicted by its grounding documents, but they fail to capture when both supporting and contradicting evidence coexist. We introduce ConflictScore, a novel metric that quantifies how well a model's response acknowledges conflicting evidence in its grounding documents. Our framework decomposes responses into atomic claims, labels each claim against each grounding document, and then aggregates these labels into two complementary measures: ConflictScore-Count (CS-C), the proportion of claims exhibiting conflicts, and ConflictScore-Ratio (CS-R), the balance between supporting and contradicting evidence. We develop ConflictBench, a benchmark covering diverse forms of conflicts such as ambiguity, contradiction, and divergent opinions, to systematically evaluate our metric. Experiments show that ConflictScore effectively detects overconfident claims across domains and can serve as a corrective feedback mechanism that improves truthfulness on TruthfulQA.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。