Reference-Free Evaluation of Reasoning in Open-Ended Question Answering

开放式问答中推理过程的无参考评估

Abstract: AI-generated answers in high-stakes domains are often fluent but difficult to verify, especially when they contain multi-step reasoning rather than a single final answer. We propose a reasoning-based, reference-free framework for auditing LLM-generated outputs. The method decomposes a generated reasoning trace into segments, labels local premise-target relations using Natural Language Inference (NLI), and organizes these relations into a hypergraph. A deterministic backward AND-OR search then assigns segment-level audit labels that indicate how each segment is grounded within the generated response.

摘要： 在高风险领域中，人工智能生成的答案往往表达流畅，但难以验证，尤其是当答案包含多步推理而非单一结论时。我们提出了一种基于推理的无参考框架，用于审计大语言模型（LLM）生成的输出。该方法将生成的推理轨迹分解为多个片段，利用自然语言推理（NLI）对局部前提与目标之间的关系进行标注，并将这些关系组织成超图。随后，通过确定性的反向“与-或”（AND-OR）搜索，为每个片段分配审计标签，以指示该片段在生成响应中的依据程度。

We evaluate the framework in two settings: deductive mathematical reasoning with Hard2Verify, and open-ended medical reasoning with UroReason, a new physician-annotated benchmark of LLM reasoning traces from real clinical cases. Across these settings, our NLI-hypergraph audit provides a more reliable reference-free evaluation signal than direct LLM-as-judge baselines. In the clinical setting, state-of-the-art LLM judges often fail to identify problematic reasoning segments, over-accepting fluent but weakly grounded responses.

我们在两种场景下对该框架进行了评估：一是使用 Hard2Verify 进行演绎数学推理评估，二是使用 UroReason 进行开放式医学推理评估——后者是一个由医生标注的、基于真实临床案例的 LLM 推理轨迹基准测试集。在这些场景中，我们的 NLI 超图审计方法比直接使用 LLM 作为评判者的基准方法提供了更可靠的无参考评估信号。在临床场景中，最先进的 LLM 评判者往往无法识别出有问题的推理片段，从而过度接受那些表达流畅但缺乏事实依据的回答。

Our results show that QA evaluation should account for how inferential relations compose across a reasoning trace, rather than relying only on final answers or LLMs as verifiers. UroReason will be made available through an API, and our code will be released as open source.

我们的研究结果表明，问答评估应考虑推理关系在整个推理轨迹中是如何构成的，而不是仅仅依赖最终答案或将 LLM 作为验证者。UroReason 将通过 API 提供，我们的代码也将开源发布。