Large language models (LLMs) often produce confident but incorrect answers in settings where abstention would be safer. Standard evaluation protocols, however, require a response and do not account for how confidence should guide decisions under different risk preferences. To address this gap, we introduce the Behavioral Alignment Score (BAS), a decision-theoretic metric for evaluating how well LLM confidence supports abstention-aware decision making. BAS is derived from an explicit answer-or-abstain utility model and aggregates realized utility across a continuum of risk thresholds, yielding a measure of decision-level reliability that depends on both the magnitude and ordering of confidence. We show theoretically that truthful confidence estimates uniquely maximize expected BAS utility, linking calibration to decision-optimal behavior. BAS is related to proper scoring rules such as log loss, but differs structurally: log loss penalizes underconfidence and overconfidence symmetrically, whereas BAS imposes an asymmetric penalty that strongly prioritizes avoiding overconfident errors. Using BAS alongside widely used metrics such as ECE and AURC, we then construct a benchmark of self-reported confidence reliability across multiple LLMs and tasks. Our results reveal substantial variation in decision-useful confidence, and while larger and more accurate models tend to achieve higher BAS, even frontier models remain prone to severe overconfidence. Importantly, models with similar ECE or AURC can exhibit very different BAS due to highly overconfident errors, highlighting limitations of standard metrics. We further show that simple interventions, such as top-$k$ confidence elicitation and post-hoc calibration, can meaningfully improve confidence reliability. Overall, our work provides both a principled metric and a comprehensive benchmark for evaluating LLM confidence reliability.

翻译：大语言模型（LLMs）在应谨慎回避的场景中常生成自信但错误的回答。然而，标准评估协议要求模型必须给出响应，并未考虑在不同风险偏好下置信度应如何引导决策。为填补这一空白，我们提出行为对齐分数（Behavioral Alignment Score, BAS），这是一种基于决策理论的指标，用于评估LLM置信度对"回避感知决策"的支持程度。BAS源于显式的"回答或回避"效用模型，通过聚合连续风险阈值上的实际效用，生成既依赖置信度大小又依赖其排序的决策层可靠性度量。我们从理论上证明：真实置信度估计能唯一最大化期望BAS效用，从而将校准性与决策最优行为相关联。BAS与对数损失等适当评分规则相关，但结构存在差异：对数损失对称惩罚过度自信和自信不足，而BAS施加非对称惩罚，优先严格避免过度自信错误。我们利用BAS与ECE、AURC等常用指标，构建了跨多个LLM和任务的自我报告置信度可靠性基准。结果表明，决策可用置信度存在显著差异：虽然更大更准确的模型往往获得更高BAS，但即使前沿模型仍易出现严重过度自信。更重要的是，由于高度过度自信的错误，具有相似ECE或AURC的模型可能表现出截然不同的BAS，这凸显了标准指标的局限性。我们还发现，简单干预措施（如top-$k$置信度提取和事后校准）能有效提升置信度可靠性。总体而言，本研究为评估LLM置信度可靠性提供了理论指标与全面基准。