causal-training

该论文提出 OneWM，探讨世界模型与动作建模在统一因果训练框架下如何相互增益，并展示了真实机器人操作（如积木堆叠、早餐准备）及模型自回归想象式推演的演示。

ForgeWM is a progressive framework that distills bidirectional video generators into efficient few-step interactive world models, supporting low-latency interaction and replay-time refinement with demonstrated improvements on Minecraft and FPS gameplay.