arXiv：cs.LG（机器学习，全量分类）· Taegeun Yang, Youngju Na, Yoonki Cho, Sung-Eui Yoon·· 13 小时前AI 评分42
CRAFT：面向 VLA 组合泛化的技能对齐方法
Same Scene, Different Task: Skill Alignment for Compositional Generalization in VLAs
AI 导读
针对 VLA 模型难以泛化到微调演示中未出现的技能组合问题，研究者提出 CRAFT，通过将已演示技能执行的监督信号迁移到反事实指令对上，实现技能表示复用。在三种 VLA 模型和两个仿真基准上，CRAFT 提升了未演示组合的成功率，同时保持已演示组合的高成功率，并在真实机器人上改善了组合泛化能力。
正文
Abstract:Vision-language-action (VLA) models often struggle to generalize to skill combinations absent from their fine-tuning demonstrations, even when every constituent skill has been demonstrated. We focus on a vision shortcut as one failure mode: during fine-tuning, visual observations can serve as a proxy for the instruction, so a policy may execute a demonstrated combination associated with similar observations rather than the instructed combination. This motivates training with counterfactual pairs formed by holding a demonstration observation fixed while changing the instruction to specify an undemonstrated combination. These pairs, however, lack corresponding demonstrated action targets. Crucially, the currently required skill has already been demonstrated, but actions from those executions cannot serve as direct targets because the same skill can require different actions across observations. We propose CRAFT, which transfers supervision from demonstrated executions of the required skill to counterfactual pairs using skill representations that can be reused across executions of the same skill. Across three VLA models and two simulation benchmarks, CRAFT improves success on undemonstrated combinations while maintaining high success on demonstrated ones; it also improves compositional generalization on a real robot. Project website: this https URL
| Comments: | 26 pages, 5 figures. Project page: this https URL |
| Subjects: | Robotics (cs.RO); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.00524 [cs.RO] |
| (or arXiv:2610.00524v1 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2610.00524 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Taegeun Yang [view email]
[v1]
Wed, 30 Sep 2026 18:13:19 UTC (21,306 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org