超越结果：面向高效智能体基准测试的双视角关系学习
Beyond Outcomes: Dual-View Relational Learning for Efficient Agent Benchmarking
- Hunyuan Team, Tencent（腾讯混元团队）
- Tsinghua University（清华大学）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
针对智能体基准评估成本高的问题，提出DualViewEval方法，联合利用结果与过程关系压缩基准，在五个基准上以20个任务实现24-40倍压缩并提升预测精度，同时揭示智能体能力差异。
AI中文摘要：
智能体基准测试的评估成本远高于传统的大语言模型基准测试。因此，基准压缩是一种自然的解决方案，然而现有方法主要对任务-模型最终得分分布中的冗余进行建模，而这在智能体评估中至关重要。为弥补这一局限，我们分析了大规模轨迹数据，识别出六个与最终智能体性能系统性相关的互补过程信号。为了从完整视角解耦智能体性能冗余，我们提出了DualViewEval，一种智能体基准压缩方法，该方法联合利用结果关系与过程关系来学习一个精确大小的最小集，并预测完整基准的得分。在五个智能体基准和五个代表性基线上，DualViewEval在所有数据集上均取得了最佳结果。仅使用20个任务，它在APEX-Agents和BFCL上实现了24倍至40倍的压缩，相较于最强竞争者将平均绝对误差（MAE）降低了14.5%至28.2%，同时在SWE-bench Verified上相对于EssenceBench将Kendall's τ提升了高达7.2%。所选的最小集还揭示了不同智能体之间的能力差异，为高效的智能体模型开发提供了紧凑且具有诊断性的反馈。
英文摘要：
Agent benchmarks are substantially more costly to evaluate than conventional LLM benchmarks. Benchmark compression is therefore a natural solution, yet existing methods primarily model redundancy in task--model final-score distributions, which is important in agentic evaluation. To address this limitation, we analyze large-scale trajectories and identify six complementary process signals that are systematically associated with final agent performance. To disentangle agent performance redundancy from a complete perspective, we propose DualViewEval, an agent benchmark compression method that jointly exploits outcome and process relations to learn an exact-size miniset and predict the full-benchmark scores. Across five agent benchmarks and five representative baselines, DualViewEval achieves the best results in all datasets. With only 20 tasks, it achieves $24\times$--$40\times$ compression on APEX-Agents and BFCL, reducing mean absolute error (MAE) by $14.5\%$--$28.2\%$ over the strongest competitors while improving Kendall's $τ$ by up to $7.2\%$ relative to EssenceBench on SWE-bench Verified. The selected minisets further reveal capability differences among different agents, providing compact and diagnostic feedback for efficient agentic model development.
