Long-WAM：扩展世界-动作模型的上下文
原标题：Paper page - Long-WAM: Scaling the Context of World-Action Models
AI 摘要
研究团队提出 Long-WAM，一个在实时控制约束下扩展因果世界-动作模型上下文的模型-系统框架。核心发现是：仅提供历史信息不等于有效利用历史，只有当视频基础模型以自回归方式预训练时，更长的历史才能带来显著收益。在 RoboCasa GR-1 上，上下文从 0.0 秒增至 19.2 秒使成功率从 63.3% 提升到 78.7%，而双向预训练初始化则无净增益。该系统通过流式观测编码、异步执行和硬件加速，可在 RTX 5090、DGX Spark 和 Jetson AGX Thor 上部署，并在 Unitree G1 和 YAM 上实现动态长程操作，动态叠杯任务成功率达 95%。
正文节选
Long-WAM: Scaling the Context of World-Action Models Abstract Real-time robot control demands enough visual history to infer motion and task progress, but processing that history can delay action. We present Long-WAM, a model-system framework for scaling the context of causal world-action models under real-time control constraints. Our central finding is that access to history is not the same as using it: longer histories pay off far more when the video foundation is pretrained autoregressively