robotics
The Robot Shortened Its Plan When the Prediction Wobbled
Denoising-path geometry raised real-world success from 53.3% to 74.4% by adapting action-chunk length.
Summary
Denoising-path geometry raised real-world success from 53.3% to 74.4% by adapting action-chunk length.
Vision-language-action policies usually commit to a fixed number of actions per query even though free motion and contact demand different feedback rates. GeoAAC reads geometric variation inside one flow-matching denoising trajectory as a reliability signal and chooses the executable prefix without retraining. Tests with GR00T N1.5 and π0.5 across LIBERO, LIBERO-Pro, RoboCasa365 and physical manipulation improved simulation results by as much as 8.7 percentage points. Average real-world success rose from 53.3% to 74.4% against fixed-horizon baselines.
Why it matters
Denoising-path geometry raised real-world success from 53.3% to 74.4% by adapting action-chunk length.
Limits and context
No additional limitation was separately recorded.
Key claims
Denoising-path geometry raised real-world success from 53.3% to 74.4% by adapting action-chunk length.
Evidence: source-2026-09-18-006
Sources
- arXiv preprint 2609.20776arXiv · primary research
Corrections
No corrections have been recorded for this story.