OpenDiscoveryTrace：评估 AI 科学家工作流的过程轨迹数据集
原标题：OpenDiscoveryTrace: Process Traces for Evaluating AI Scientist Workflows
AI 摘要
研究者发布 OpenDiscoveryTrace，首个公开的 AI 科学智能体过程轨迹数据集，包含 558 条完整轨迹，覆盖 7 个模型和 124 个科学任务。每条轨迹记录 9 字段的逐步推理过程（思维、工具调用、观察、错误、修正触发、置信度等），而非仅最终输出。初步分析显示，三个前沿模型成功率相近（84–89%），但错误类型差异显著：Claude Opus 4.6 多为工具误用（66.7%），GPT-5.4 多为推理错误（83.6%），表明仅评估输出的基准会遗漏过程层面的行为差异。数据集、轨迹模式、智能体框架和基准定义以 CC-BY-4.0 协议公开。
正文节选
OpenDiscoveryTrace: Process Traces for Evaluating AI Scientist Workflows Abstract Existing benchmarks for autonomous AI scientists evaluate only final outputs—generated code, hypotheses, or papers—yet discard the reasoning process by which those outputs were obtained. This makes it impossible to audit scientific methodology, diagnose failure modes, or distinguish systematic reasoning from fortunate guessing. We present OpenDiscoveryTrace, a public dataset of 558 complete AI scientific agent traj