用隐空间回放实现持续学习，减少遗忘并提升多任务表现。
Lifelong Imitation Learning with Multimodal Latent Replay and Incremental Adjustment
- 在多模态隐空间中存储视觉、语言和状态信息，替代传统经验回放。
- 在LIBERO基准上提升AUC 10-17点，遗忘率降低65%。
- 适合长期机器人学习场景，尤其受限于内存与数据的部署环境。
我们提出一种终身模仿学习框架，在真实内存与数据限制下，实现序列任务上的持续策略优化。该方法摒弃传统经验回放，完全在多模态隐空间中操作，将视觉、语言和机器人状态信息的紧凑表示进行存储与复用，以支持后续学习。为增强适应稳定性，引入增量特征调整机制，通过角度边界约束正则化任务嵌入的演化过程，保持任务间的区分性。在LIBERO基准上，该方法达到新最佳性能，AUC提升10-17点，遗忘率相比先前最优方法降低高达65%。消融实验验证各组件有效性，各项替代策略均表现逊色。代码已开源：https://github.com/yfqi/lifelong_mlr_ifa。
原文摘要 · Abstract (English)
We introduce a lifelong imitation learning framework that enables continual policy refinement across sequential tasks under realistic memory and data constraints. Our approach departs from conventional experience replay by operating entirely in a multimodal latent space, where compact representations of visual, linguistic, and robot's state information are stored and reused to support future learning. To further stabilize adaptation, we introduce an incremental feature adjustment mechanism that regularizes the evolution of task embeddings through an angular margin constraint, preserving inter-task distinctiveness. Our method establishes a new state of the art in the LIBERO benchmarks, achieving 10-17 point gains in AUC and up to 65% less forgetting compared to previous leading methods. Ablation studies confirm the effectiveness of each component, showing consistent gains over alternative strategies. The code is available at: https://github.com/yfqi/lifelong_mlr_ifa.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。