Dejavu：迈向具身智能的经验反馈学习
Dejavu: Towards Experience Feedback Learning for Embodied Intelligence
AI总结：
本文提出Dejavu框架，通过经验反馈网络增强冻结的视觉-语言-动作策略，提升具身智能任务的适应性、鲁棒性和成功率。
摘要：
具身智能代理面临根本性限制：一旦部署到现实环境，难以轻易获取新知识以提升任务表现。本文提出Dejavu，一种通用后部署学习框架，通过经验反馈网络（EFN）将检索到的执行记忆与冻结的视觉-语言-动作（VLA）策略结合。EFN识别上下文相关的先前行动经验，并根据检索到的指导条件行动预测。我们用强化学习和语义相似性奖励训练EFN，促使预测动作与过去行为在当前观察下对齐。在部署过程中，EFN持续扩展其记忆，使代理能够表现出『从经验学习』。在多样化的具身任务中，EFN在适应性、鲁棒性和成功率方面优于冻结基线。我们的项目页面是https://dejavu2025.github.io/。
英文摘要：
Embodied agents face a fundamental limitation: once deployed in real-world environments, they cannot easily acquire new knowledge to improve task performance. In this paper, we propose Dejavu, a general post-deployment learning framework that augments a frozen Vision-Language-Action (VLA) policy with retrieved execution memories through an Experience Feedback Network (EFN). EFN identifies contextually relevant prior action experiences and conditions action prediction on the retrieved guidance. We train EFN with reinforcement learning and semantic similarity rewards, encouraging the predicted actions to align with past behaviors under the current observation. During deployment, EFN continually expands its memory with new trajectories, enabling the agent to exhibit ``learning from experience.'' Experiments across diverse embodied tasks show that EFN improves adaptability, robustness, and success rates over frozen baselines. Our Project Page is https://dejavu2025.github.io/.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
