给大模型生成加风险控制，确保幻觉出现频率可控。
Chance-Constrained Inference for Hallucination Risk Control in Large Language Models
- 将幻觉建模为随机约束违规，用概率上限控制生成错误频率。
- 实验证明能可靠控制幻觉率，且在重复使用中保持安全。
- 适合对生成可靠性要求高的场景，如医疗、法律问答。
大型语言模型生成结果具有随机性，可能产生流畅但不正确的回应，包括事实性幻觉。现有缓解策略虽降低了平均错误率，却无法在重复使用中明确控制此类失败的频率。本文将推理视为部署阶段的风险控制问题，提出‘机会约束推理’（chance-constrained inference），直接限制被接受生成结果中幻觉发生的概率。我们将幻觉建模为随机约束违反，并证明基于置信度的选择性预测通常不蕴含概率风险保证。为高效实施机会约束，我们提出一种顺序的、随时有效的推理过程，通过有限样本自适应地认证可行性或不可行性，避免保守的固定样本边界。在受NaturalQuestions启发及受控多跳问答任务上的实验表明，该方法能实现可靠的風險控制，早期识别内在不可行输入，并在重复使用下安全组合生成，而基于置信度的基线方法无法提供一致保障。
原文摘要 · Abstract (English)
Large language models generate outputs stochastically and may produce fluent but invalid responses, including factual hallucinations. Existing mitigation strategies reduce average error rates but do not provide explicit control over the \emph{frequency} of such failures under repeated use. We formulate inference as a deployment-time risk control problem and introduce \emph{chance-constrained inference}, which directly bounds the probability of hallucinations among accepted generations. Hallucinations are modeled as stochastic constraint violations, and we show that confidence-based selective prediction does not, in general, imply probabilistic risk guarantees. To enforce chance constraints efficiently, we propose a sequential, anytime-valid inference procedure that adaptively certifies feasibility or infeasibility using finite samples, avoiding conservative fixed-sample bounds. Experiments on questions inspired by NaturalQuestions and controlled multi-hop question answering demonstrate reliable risk control, early detection of intrinsically infeasible inputs, and safe composition under repeated use, while confidence-based baselines fail to provide consistent guarantees.
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。