测试大模型推理过程是否真实可靠，发现近半数输出存在逻辑不一致。
RFEval: Benchmarking Reasoning Faithfulness under Counterfactual Reasoning Intervention in Large Reasoning Models
- 通过反事实干预测试推理与答案的因果关联性
- 49.7%的输出存在不忠实地推理，数学代码类任务最差
- 准确率不能代表推理可信度，需单独评估推理结构
大型推理模型虽表现优异，但常生成看似合理实则不符真实决策过程的解释，损害可靠性。本文提出推理忠实性的形式化框架，包含立场一致性（推理与答案逻辑连贯）和因果影响（干预推理内容可改变答案）两个可检验条件，且与准确率解耦。为此构建了涵盖7个任务、共7,186个实例的RFEval基准，通过输出级反事实干预检测忠实性。对12个开源模型评估显示，49.7%的输出存在不忠实，主要源于立场不一致。失败集中于数学、代码等脆弱收敛领域，且与后训练策略相关：在监督微调基础上添加当前强化学习目标反而降低忠实性，即使准确率不变。关键发现是，准确率既非充分也非可靠的忠实度代理指标——控制模型与任务后，两者关联极弱且统计不显著。本研究建立了审计大模型可靠性的严谨方法，表明可信AI需同时优化结果正确性与推理过程结构完整性。代码与数据集详见：https://aidaslab.github.io/RFEval/
原文摘要 · Abstract (English)
Large Reasoning Models (LRMs) exhibit strong performance, yet often produce rationales that sound plausible but fail to reflect their true decision process, undermining reliability and trust. We introduce a formal framework for reasoning faithfulness, defined by two testable conditions: stance consistency (a coherent stance linking reasoning to answer) and causal influence (the stated reasoning causally drives the answer under output-level interventions), explicitly decoupled from accuracy. To operationalize this, we present RFEval, a benchmark of 7,186 instances across seven tasks that probes faithfulness via controlled, output-level counterfactual interventions. Evaluating twelve open-source LRMs, we find unfaithfulness in 49.7% of outputs, predominantly from stance inconsistency. Failures are concentrated in brittle, convergent domains such as math and code, and correlate more with post-training regimes than with scale: within-family ablations indicate that adding current RL-style objectives on top of supervised fine-tuning can reduce reasoning faithfulness, even when accuracy is maintained. Crucially, accuracy is neither a sufficient nor a reliable proxy for faithfulness: once controlling for model and task, the accuracy-faithfulness link is weak and statistically insignificant. Our work establishes a rigorous methodology for auditing LRM reliability and shows that trustworthy AI requires optimizing not only for correct outcomes but also for the structural integrity of the reasoning process. Our code and dataset can be found at project page: https://aidaslab.github.io/RFEval/
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。