CARE: Experience-Guided Atomic Corrective Execution for Vision-Language-Action Policies

类型：arxiv
标识：2609.24118
链接：https://arxiv.org/abs/2609.24118
主分类：multimodal
形态：method
TLDR：Vision-Language-Action (VLA) policies achieve strong performance in robotic manipulation but remain brittle once execution deviates from nominal trajectories. We propose CARE (Corrective Atomic Robotic Execution), a framework that improves recovery by learning from failures encountered during execution. Instead of generating corrective data from manually designed or random perturbations, CARE collects failed rollouts, models stage-conditioned post-failure deviations, and uses the resulting empirical distributions to synthesize representative failure states and corrective demonstrations. At inf
待LLM分类：否
来源文件：
/inbox/tom/_candidates/2026-09-22-agent-rag-longcontext-candidates.json

2026-06-27

笔记 xHC: Expanded Hyper-Connections · 精读与批判(2026-07-21 09:50)

2026-07-21

仓库 EvolvingLMMs-Lab/lmms-eval

2026-08-29

笔记精读：Efficient Inference for Large Vision-Language Models: Bottlenecks, Techniques, and Prospects

2026-07-12

笔记 2026-09-18 在线视频 LLM × 多模态对齐评估（轻量精读）

2026-09-18

笔记 2026-07-04 上午多模态对照精读 · Seeker（文本→像素的长上下文 MLLM）

2026-07-04

arXiv

2609.24118

TLDR

Vision-Language-Action (VLA) policies achieve strong performance in robotic manipulation but remain brittle once execution deviates from nominal trajectories. We propose CARE (Corrective Atomic Robotic Execution), a framework that improves recovery by learning from failures encountered during execution. Instead of generating corrective data from manually designed or random perturbations, CARE collects failed rollouts, models stage-conditioned post-failure deviations, and uses the resulting empirical distributions to synthesize representative failure states and corrective demonstrations. At inf

arXiv ↗ 中文全文

CARE: Experience-Guided Atomic Corrective Execution for Vision-Language-Action Policies

相关 · 按语义匹配