Echo denoises stop and failure signals inside vision-language action policies
Original title: Echo: Coupled Action and Task Status Signal Denoising for Vision Language Action Models
Echo learns task status without frame-level progress labels and uses those signals to condition later action denoising.
Echo adds stop and failure signals as two extra dimensions of each action chunk and denoises them with robot actions inside one vision-language action model. Successful, failed, and recovery trajectories supply supervision without frame-level progress annotations. Signals are denoised faster than actions so later action steps can condition on cleaner status estimates. Real-world manipulation experiments show accurate status prediction and improved policy performance.
- Stop and failure signals distinguish ongoing execution, success, and failure.
- The method avoids frame-level progress annotations and external monitoring models.
- Status signals are denoised faster than the accompanying robot actions.
- The paper reports real-world manipulation tests but names no robot or dataset.
From the original article
1 min read at huggingface.co
Recognizing task completion and failure enables robots to stop upon success and initiate recovery or replanning after failure. Existing methods augment Vision Language Action (VLA) policies with task status signal prediction to support these execution decisions. However, learning and integrating these signals can require fine-grained supervision, external monitoring models, or complex architectural designs.
By Yufeng Liu, Bohao Fan, Juntu Zhao · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.