Towards Robust Long-Lived Agents: Benchmarking Experience Lifecycles in Evolving Environments

查看原文 / View Original 编辑 / Edit

Computer Science > Artificial Intelligence arXiv:2611.05450 (cs) [Submitted on 3 Nov 2026] Title: Towards Robust Long-Lived Agents: Benchmarking Experience Lifecycles in Evolving Environments Authors: Shijue Huang, Hangyu Guo, Hengheng Zhang, Guo An, Zhitao He, Qi Tian, Yiren Feng.

计算机科学 > 人工智能 arXiv:2611.05450 (cs) [提交于 2026 年 11 月 3 日] 标题：迈向稳健的长寿命智能体：演化环境中的经验生命周期基准测试作者：Shijue Huang, Hangyu Guo, Hengheng Zhang, Guo An, Zhitao He, Qi Tian, Yiren Feng。

Abstract: As agents become more capable, they are moving beyond one-off tasks toward sustained assistance across successive user goals. Such assistance requires agents to discover how unfamiliar environments work, reuse what they learn across tasks, and keep that experience current as environments evolve.

摘要：随着智能体能力不断增强，它们正从处理一次性任务转向在连续的用户目标中提供持续的辅助。这种辅助要求智能体能够探索陌生环境的运作方式，在不同任务间复用所学知识，并随着环境的演变保持经验的实时更新。

Existing benchmarks increasingly study experience reuse and adaptation, but provide limited insight into which parts of accumulated experience agents revise, retain, or lose. To address this gap, we introduce ReScopeBench, a benchmark for the experience lifecycle across 18 interactive environments.

现有的基准测试越来越多地研究经验的复用与适应，但对于智能体在积累的经验中哪些部分被修订、保留或丢失，提供的见解仍然有限。为了填补这一空白，我们引入了 ReScopeBench，这是一个涵盖 18 个交互式环境的经验生命周期基准测试。

It distinguishes structural experience about how an environment is organized and accessed from rule experience about reusable principles and constraints that guide decisions across tasks. To study how experience is acquired, reused, and revised, we propose an experience-first construction approach.

它将关于环境如何组织和访问的“结构经验”，与指导跨任务决策的可复用原则和约束的“规则经验”区分开来。为了研究经验是如何被获取、复用和修订的，我们提出了一种“经验优先”的构建方法。

It builds task streams around shared experience and introduces silent environmental changes. We evaluate task performance and track changes in saved experience throughout each stream.

它围绕共享经验构建任务流，并引入了隐性的环境变化。我们在整个任务流中评估任务性能，并追踪已保存经验的变化。

Across 20 models, we find that retained experience improves task performance, but agents struggle to keep it current as environments change. Even when the required experience is correctly recorded, agents do not always complete tasks successfully. These findings identify selective maintenance and effective use as complementary challenges for long-lived agents.

通过对 20 个模型的研究，我们发现保留的经验确实能提升任务性能，但当环境发生变化时，智能体难以保持经验的实时性。即使所需的经验被正确记录，智能体也并不总能成功完成任务。这些发现表明，选择性维护和有效利用是长寿命智能体面临的互补性挑战。