VLA 有多少是 kNN？检索复用演示在 LIBERO 上达 91.7%
How Much of Your VLA Is kNN
AI 导读
一项 arXiv 研究表明，仅检索最近演示并复用其动作块、不做任何策略训练，就能在 LIBERO 仿真基准上取得 91.7% 成功率，而 VLA 模型在该基准上已超过 95%。
正文 · 原文
Abstract:Vision-language-action (VLA) models now exceed 95% success on the LIBERO simulation benchmark, yet it remains unclear how much of this performance requires generalization beyond the training demonstrations. In this work, we show that retrieving nearby demonstrations and reusing their action chunks achieves a success rate of 91.7 % on LIBERO without policy training. Under LIBERO-Plus perturbations, retrieval shows the smallest drop of all evaluated policies. Across nine publicly available VLAs, fine-tuning mostly makes the vision-language backbone a better retrieval key, and the actions of the VLAs remain close to the demonstrations. In addition, we introduce Local Action Consistency (LAC), an offline metric that measures whether nearby observations in representation space are also associated with similar future actions. We show that LAC tracks the success rate of retrieval and can guide the training of a retrieval key. Our evaluation and source code are available at this https URL.
| Comments: | 21 pages, 10 figures, accepted to the CoRL 2026 G2A/TAG workshop |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2611.13160 [cs.RO] |
| (or arXiv:2611.13160v1 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2611.13160 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Damien LaRocque [view email]
[v1]
Mon, 9 Nov 2026 19:44:58 UTC (2,159 KB)
来源：arXiv cs.AI · arxiv.org