用混合效用最小贝叶斯风险降低大模型幻觉，提升企业级应用可靠性。
Reducing Hallucination in Enterprise AI Workflows via Hybrid Utility Minimum Bayes Risk (HUMBR)
- 将幻觉抑制建模为最小贝叶斯风险问题，结合语义与词法相似度找共识。
- 在真实生产数据上，81%的生成建议优于人工基准，关键召回错误几乎消除。
- 适用于法律、风控等高风险场景，适合关注可信AI落地的从业者。
尽管大语言模型推动自动化，但在法律、风险管理及隐私合规等高风险企业流程中仍需高度谨慎。对Meta等机构而言，此类流程中一个幻觉条款可能带来重大后果。本文将幻觉缓解视为最小贝叶斯风险（MBR）问题，提出混合效用最小贝叶斯风险（HUMBR）框架，通过融合语义嵌入相似性与词汇精确度，在无真实参考情况下识别共识，并推导出严格的误差界。理论分析结合对TruthfulQA和LegalBench等公开基准以及Meta生产环境真实数据的全面实证评估显示，MBR显著优于通用自一致性方法。实验表明，该流水线81%的建议优于人工标注基准，关键召回失败近乎归零。
原文摘要 · Abstract (English)
Although LLMs drive automation, it is critical to ensure immense consideration for high-stakes enterprise workflows such as those involving legal matters, risk management, and privacy compliance. For Meta, and other organizations like ours, a single hallucinated clause in such high stakes workflows risks material consequences. We show that by framing hallucination mitigation as a Minimum Bayes Risk (MBR) problem, we can dramatically reduce this risk. Specifically, we introduce a Hybrid Utility MBR (HUMBR) framework that synthesizes semantic embedding similarity with lexical precision to identify consensus without ground-truth references, for which we derive rigorous error bounds. We complement this theoretical analysis with a comprehensive empirical evaluation on widely-used public benchmark suites (TruthfulQA and LegalBench) and also real world data from Meta production deployment. The results from our empirical study show that MBR significantly outperforms standard Universal Self-Consistency. Notably, 81% of the pipeline's suggestions were preferred over human-crafted ground truth, and critical recall failures were virtually eliminated.
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。