论文/arXiv 2610.13021·19 天前

EmbodiedCLBench：评估自进化具身智能体的持续学习能力

提出EmbodiedCLBench，在具身仿真中以组合泛化评估自进化智能体的持续学习，并用多种智能体框架和学习方法验证。

AI 整理

• 仿真与训练
• 数据与采集
• 评测与安全

原文链接：EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents（arXiv 2610.13021 · 发布日期：2026-10-09）

EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents

提出EmbodiedCLBench，在具身仿真中以组合泛化评估自进化智能体的持续学习，并用多种智能体框架和学习方法验证。

核心要点

• 现有持续学习研究多局限于非具身场景，缺少面向具身环境的评估。
• 基准要求智能体先学基础任务，再重组已获技能解决未见高级任务。
• 多种智能体框架和学习方法实验显示，基础任务经验能稳定提升高级任务表现。
• 改进未随学习样本数量相应扩大，代码和数据集已公开。

验证范围与局限：验证范围为具身仿真环境，实验覆盖多种智能体框架和学习方法。

英文摘要 (Abstract)

Continual learning requires models to accumulate past experience and transfer it to new tasks. Despite substantial research on continual learning, existing work remains largely confined to non-embodied settings, without extending to embodied environments. To address this gap, we introduce EmbodiedCLBench, the first benchmark in embodied simulation for evaluating the continual learning capability of self-evolving agents. With compositional generalization as the core evaluation principle, our benchmark evaluates whether agents can first learn from basic tasks and then recombine the acquired skills to solve complex, unseen advanced tasks. Based on EmbodiedCLBench, we conduct extensive experiments across multiple agent harnesses and learning methods, revealing consistent limitations and distinctive behaviors. Our results show that while advanced tasks can be hardly solved under zero-shot setting, the experience from basic tasks enables consistent improvement. Such benefit from the experience remains stable even if the task complexity increases. However, this improvement does not scale accordingly with the number of learning samples, and current agents cannot extract meaningful information from additional experience. Collectively, our findings highlight continual learning as a fundamental yet underexplored capability, and our benchmark offers valuable resources for designing effective self-evolving agents in embodied environments. Our code and dataset are available at https://pku-value-lab.github.io/EmbodiedCLBench-homepage.

作者：Jie Huang, Yanan Chen, Ruixun Liu, Kaichen He, Xinyi He, Rui Huang, Zilong Zheng, Zhenliang Zhang, Yiwu Zhong

来源

本页事实对应的公开页面，以原文为准。

1. arXiv 2610.13021：EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents访问 2026-10-21