Breaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Models
- 类型：arxiv
- 标识：2609.12641
- 链接：https://arxiv.org/abs/2609.12641
- 主分类：engineering
- 形态：method
- TLDR：Robot foundation models achieve strong in-distribution performance but often degrade under visual distribution shifts. When learning to generate actions from pretrained visual representations, models may exploit task-irrelevant visual cues that correlate with demonstrated actions within the training distribution. Such vision-action shortcuts can undermine generalization when these correlations change under distribution shifts. Mitigating these shortcuts requires constraining how visual information is used for action generation while preserving task-relevant spatial information. We propose Late
- 副分类：multimodal
- 待LLM分类：否
- 来源文件：
- /inbox/tom/_candidates/2026-09-14-agent-rag-longcontext-candidates.json