arXiv cs.AI· Carlota Par\'es-Morlans, Nils Kuhn, Y. Isabel Liu, Alberta Longhini, Jeannette Bohg·· 12 小时前AI 评分38
研究揭示 VLA 模型为何在接触密集型任务中失败，并提出 FACT 修复方案
Demystifying When and Why VLAs Fail in Contact-Rich Tasks and How to Fix Them
AI 导读
研究识别出 Vision-Language-Action 模型在接触密集型操作中的两种失败模式：精度失败（难以产生接触所需的微小精确修正动作）和力失败（无法响应接触力）。为此提出 FACT，在后训练阶段提高低噪声采样以改善精细动作预测，并引入结构化力条件引导动作生成。在五项接触密集型任务、近 2500 次真实世界 rollout 中，FACT 平均成功率达 66%，最佳先前基线为 41%。
正文 · 原文
Abstract:We address the problem of understanding when and why Vision-Language-Action models struggle with contact-rich manipulation tasks that require precise physical interaction. Prior work has primarily focused on addressing contact failures through force-augmented architectures and training-time regularizers, yet the factors contributing to these failures remain underexplored. We identify two distinct failure modes underlying this gap. Precision failures occur when policies struggle to produce the small, accurate corrective actions required during contact and force failures occur when policies fail to respond to contact forces. To address these failures, we propose FACT. During post-training, FACT increases low noise sampling to improve fine action prediction and incorporates structured force conditioning to guide action generation using contact state and force history. FACT achieves an average success rate of 66% across five contact-rich tasks against 41% for the best prior baseline, in an evaluation spanning almost 2,500 real-world rollouts.
| Comments: | 17 pages, Spotlight at CoRL 2026 |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.01402 [cs.RO] |
| (or arXiv:2608.01402v2 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2608.01402 arXiv-issued DOI via DataCite |
Submission history
From: Carlota Parés Morlans [view email]
[v1]
Sun, 2 Aug 2026 17:34:19 UTC (5,032 KB)
[v2]
Fri, 16 Oct 2026 06:10:02 UTC (39,339 KB)
来源：arXiv cs.AI · arxiv.org