外观
论文arXiv 2610.2075014 天前
机器人基础模型的因果分析带来效率提升
原文链接：Causal Analysis of Robot Foundation Models Enables Efficiency Gains（arXiv 2610.20750 · 发布日期：2026-10-14）
Causal Analysis of Robot Foundation Models Enables Efficiency Gains
研究用激活修补分析三种VLA模型中图像与指令影响动作的因果路径，并在评估与真机部署中验证效率提升。
核心要点
- VLA如何将视觉语言表示转化为动作仍不清楚，限制对动作生成机制的理解。
- 作者通过激活修补分析图像与指令到动作的因果路径，定位action-conditioning interface。
- 结果显示目标名词早中层决定选物，文本中空间信息在中后层重定向运动。
- 仅用MolmoBot的3/36层进行动作专家条件化，FLOPs最高降33%，成功率相当。
验证范围与局限：验证范围包括三种VLA模型的机制分析、MolmoBot的3,000个评估回合，以及π0.5真机部署。
英文摘要 (Abstract)
Vision-language-action (VLA) models are a promising approach to general purpose robot manipulation. While the internal computation of vision-language models, which compose the backbone of VLAs, has been studied extensively, less is known about how their perceptual and semantic representations are transformed into action in VLA models. We study this question mechanistically in three frontier VLA models ($π_{0.5}$, MolmoAct2, and MolmoBot) by analyzing the causal pathways through which the input image and the instruction shape actions. Across the three models, activation patching shows that causally relevant information influences the action expert (AE) through a narrow band of mid-to-late cross-attention layers, which we term the action-conditioning interface. The content of this information changes with token and layer within the model. Early-to-mid layer representations of the target noun token determine which object is selected, whereas mid-to-late layers carry the spatial information in text tokens that redirects the movement when patched. Where spatial information is stored in prompt tokens various based on the architecture. In MolmoAct2 the model forms task-conditioned representations at the text positions that follow the instruction, whereas $π_{0.5}$ additionally stores this information at image token positions. We show that these findings can be translated directly into improving model efficiency. We find that restricting MolmoBot's action expert conditioning to only 3 out of 36 layers reduces inference FLOPs by up to $33%$ while maintaining comparable task success with the default model across 3,000 evaluation episodes ($46.7%$ vs. $46.9%$), and the same principle carries over to real-robot deployment of $π_{0.5}$.
作者：Aiden Swann, Mathilde Papillon, Mozes Jacobs, Siddharth Boppana, Tal Haklay, Eric Bigelow, Andrew Lee, Conor Watts, Thomas Icard, Mac Schwager
来源
本页事实对应的公开页面，以原文为准。