Dejavu：面向具身智能的经验反馈学习
Dejavu: Towards Experience Feedback Learning for Embodied Intelligence
对做 VLA/具身策略的视觉研究者，这篇给出一种不改动冻结策略权重的部署后增强范式：用检索到的历史动作经验作为条件输入，并通过强化学习加语义相似度奖励训练一个轻量 EFN，可直接套用到现有 VLA 上提升成功率与鲁棒性。
原文依据 augments a frozen Vision-Language-Action (VLA) policy with retrieved execution memories through an Experience Feedback Network (EFN)
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论具身智能体面临一个根本性限制：一旦部署到真实环境中，它们便难以获取新知识以提升任务表现。本文提出 Dejavu，一个通用的部署后学习框架，它通过经验反馈网络（EFN）以检索到的执行记忆来增强一个冻结的视觉-语言-动作（VLA）策略。EFN 识别上下文相关的先验动作经验，并以检索到的指导为条件进行动作预测。我们使用强化学习与语义相似度奖励训练 EFN，促使预测动作在当前观测下与过往行为对齐。在部署期间，EFN 持续用新的轨迹扩展其记忆，使智能体展现出“从经验中学习”的能力。在多种具身任务上的实验表明，EFN 相较冻结基线提升了适应性、鲁棒性与成功率。项目主页为 https://dejavu2025.github.io/。
英文原文摘要 984 字符
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/.