测试大模型在设备维修中应对非标准操作的能力，发现其常选看似合理实则错误的步骤。
DiagFlowBench: Evaluating How Language Models Handle Off-Procedure Inputs in Grounded Diagnostic Dialogue
- 构建50个工业诊断流程图转为1676轮对话，对比合规与非合规语句
- 十款模型在偏离流程时普遍选择看似真实却错误的步骤，而非拒绝回答
- 揭示现有对话系统在真实场景下的关键缺陷，适合检修系统开发者参考
语言模型越来越多地用于维护操作的辅助决策。为防止幻觉，现有方法将模型锚定在程序文档上，限制其只能遵循预定流程。然而实际操作中，工作人员可能偏离流程，要求模型识别非脚本化的非流程语句。当前基准很少关注这一能力。我们提出DiagFlowBench，一个包含50个来自消费类制造商的工业诊断流程图的数据集，转化为1,676轮多轮对话，对比合规与非流程语句。评估十款商业及开源模型发现，其拒答率差异显著，多数模型倾向于选择看似真实但上下文不合适的步骤，而非编造事实。这种看似合理却错误的建议暴露了基于流程系统的深层脆弱性。代码与数据已公开于https://github.com/guille-gil/DiagFlowBench。
原文摘要 · Abstract (English)
Language models increasingly serve as advisory systems in maintenance operations. To prevent hallucination, established approaches ground these models in procedural documentation, constraining them to prescribed sequences. In practice, however, operators may stray from these steps, requiring models to recognise unscripted, off-procedure utterances. Current benchmarks rarely prioritise this capability. We introduce DiagFlowBench, a dataset of 50 industrial diagnostic flowcharts from a consumer manufacturer converted into 1,676 multi-turn conversations that contrast compliant with off-procedure utterances. Evaluating a panel of ten commercial and open-weight models reveals high variability in abstention rates, with models often selecting a real but contextually inadequate step rather than fabricating facts. The inherent plausibility and authority of this mapped but wrong advice exposes a challenging vulnerability for grounding systems. Code and data are released at https://github.com/guille-gil/DiagFlowBench.
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。