Robota-Embed: Unified VLA Embeddings for Robot In-Context Learning and Evolution
Robota-Embed: Unified VLA Embeddings for Robot In-Context Learning and Evolution
Robota-Embed is the first vision-language-action (VLA) embedding model that jointly encodes task, execution stage, and future continuation for robots. It achieves 47.33% clip-level and 81.77% episode-level R@1, outperforming action-free Gemini Embedding 2. The model supports training-free and training-based policy conditioning, improving success rates in simulated and real-world tasks, and enables task-local context evolution with reduced policy token usage.
This paper introduces Robota-Embed, a novel VLA embedding model that encodes vision, language, and action modalities for robot in-context learning and evolution. It demonstrates significant improvements over existing models in retrieval accuracy and task success rates across simulated and real-world environments.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org