arXiv：cs.LG· Quanquan Peng, Yutong Liang, Rui Yan, Nicklas Hansen, Xiaolong Wang·· 16 小时前
FACT：面向世界-动作模型的失败感知因果训练
FACT: Failure-Aware Causal Training for World-Action Models
AI 导读
FACT 是一种因果世界-动作模型（WAM），以已执行动作为条件预测未来视频与任务进度，使失败轨迹也能监督动作后果，而非被丢弃。该失败感知训练让进度预测器同时了解成功与失败结果，并可在推理时为候选动作打分；在仿真与真实双臂操作任务上，FACT 优于多项现有基线，引入失败数据后性能提升，并减少了失败动作下偏向成功的未来幻觉。
正文
Abstract:Recent world-action models (WAMs) show that co-training policies with future prediction can provide physical priors for action generation. Building on the future-prediction ability of video models, many WAMs generate future videos and recover actions with inverse-dynamics models, or use these predicted videos as goal conditions for action generation. In both cases, the world model is trained mostly on successful demonstrations and has little reason to predict the consequences of bad actions. We introduce FACT, a causal World-Action Model that predicts future video and task progress conditioned on the executed action. This action-conditioned interface allows failure rollouts to supervise action consequences, turning bad actions into valid future targets rather than being discarded. Failure-aware training makes the progress predictor aware of both successful and failed action outcomes, which can optionally be used to score sampled action candidates at inference. Extensive experiments on simulation and real-world bimanual manipulation tasks show that FACT outperforms many existing baselines, improves as failure data are incorporated into training, and reduces success-biased future hallucination under bad actions. See more details at this https URL
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2608.10232 [cs.RO] |
| (or arXiv:2608.10232v2 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2608.10232 arXiv-issued DOI via DataCite |
Submission history
From: Quanquan Peng [view email]
[v1]
Mon, 10 Aug 2026 21:10:46 UTC (5,022 KB)
[v2]
Tue, 13 Oct 2026 20:03:30 UTC (5,022 KB)
来源：arXiv：cs.LG · arxiv.org