跳到正文
arXiv cs.AI· Fernando Diaz·· 2 小时前AI 评分39

离线偏好式轨迹评估:用轨迹偏好对比降低智能体评测平局

Offline Preference-Based Trajectory Evaluation

AI 导读

研究提出基于偏好的轨迹评估方法,直接用时间回报曲线的时间偏好两两比较轨迹,取代只看最终成败的二值指标。在多种智能体与交互式 benchmark 上,成败指标约 75% 的实例出现平局,新方法降至约 35%,提升区分力、排名稳定性与数据效率。结果表明 benchmark 饱和可能源于评测指标设计本身。

正文 · 原文

View PDF HTML (experimental)

Abstract:Offline evaluation of agentic systems often collapses trajectories into a binary terminal success value, discarding information about partial progress. This, in turn, results in widespread ties that create statistical inefficiency by reducing effective sample size and weakening the ability to distinguish systems. We propose preference-based trajectory evaluation, an evaluation measure that compares pairs of trajectories directly by using temporal preferences between time-to-return profiles. We find that, across diverse agentic and interactive benchmarks, success-based metrics result in tied comparisons on roughly 75% of instances, while trajectory-aware preferences reduce ties to roughly 35%, improving discriminative power, ranking stability, and data efficiency. Our results suggest that benchmark saturation, often portrayed as the result of poor data collection or problem difficulty, may also be explained by the design of the evaluation measure.
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
Cite as: arXiv:2606.17541 [cs.LG]
  (or arXiv:2606.17541v2 [cs.LG] for this version)
  https://doi.org/10.48550/arXiv.2606.17541

arXiv-issued DOI via DataCite

Submission history

From: Fernando Diaz [view email]
[v1] Tue, 16 Jun 2026 05:42:19 UTC (449 KB)
[v2] Mon, 16 Nov 2026 19:51:01 UTC (464 KB)

来源:arXiv cs.AI · arxiv.org