Summary
arXiv:2611.19709v1 Announce Type: new Abstract: Flow-matching robot policies generate each action chunk through repeated network evaluations, one per denoising step. Cross-cycle caches such as C$^3$ache reuse computation from earlier replanning calls, but whether they beat simply taking fewer steps is unclear. We compare the two in Fast-WAM, a 6B-parameter world action model, on LIBERO-Goal. Two uncached steps keep the success of ten: on all 500 initial states they are non-inferior within two percentage points, for one checkpoint and noise seed. Under PyTorch's default compile mode, they also cut a replanning call on an RTX A6000 from 202 to 92 ms. This leaves little room for caching. A cache that keeps two fresh steps cannot be faster than two uncached steps: every call still encodes the images, runs the video expert and computes those two steps. Caching the first of two steps, refreshed every fourth call, saves at most 11--14% of a call by our cost model. Cached policies are also fra…
Original text(via arXiv cs.RO)
Computer Science > Robotics
Title:Cache or Denoise Less? Residual Reuse in a Flow-Matching World Action Model
Abstract:Flow-matching robot policies generate each action chunk through repeated network evaluations, one per denoising step. Cross-cycle caches such as C$^3$ache reuse computation from earlier replanning calls, but whether they beat simply taking fewer steps is unclear. We compare the two in Fast-WAM, a 6B-parameter world action model, on LIBERO-Goal. Two uncached steps keep the success of ten: on all 500 initial states they are non-inferior within two percentage points, for one checkpoint and noise seed. Under PyTorch's default compile mode, they also cut a replanning call on an RTX A6000 from 202 to 92 ms. This leaves little room for caching. A cache that keeps two fresh steps cannot be faster than two uncached steps: every call still encodes the images, runs the video expert and computes those two steps. Caching the first of two steps, refreshed every fourth call, saves at most 11--14% of a call by our cost model. Cached policies are also fragile. Their losses tracked the share of the flow path the cached steps cover: under Fast-WAM's current time shift, C$^3$ache's eight cached steps cover 80% of the path instead of 44%, and the cached policy succeeds on 63% of states against 95% uncached. Resampling the noise on every call cost caches covering 60% or more of the path 14--40 points. These results argue for fewer-step baselines and for testing reuse under the deployed sampler and noise settings. We release our evaluation harness, logs and a checklist.
| Comments: | Honorable Mention at the CoRL 2026 Workshop on Efficient Foundation Models for Real-Time Embodied AI (EEAI). Code and data available at this https URL |
|---|
| Subjects: | Robotics (cs.RO) ; Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| MSC classes: | 68T40, 68T07 |
| ACM classes: | I.2.9; I.2.6 |
| Cite as: | arXiv:2611.19709 [cs.RO] |
| (or arXiv:2611.19709v1 [cs.RO] for this version) |
| https://doi.org/10.48550/arXiv.2611.19709 arXiv-issued DOI via DataCite (pending registration) |
Submission history
The text above was extracted automatically from the source page and may miss images, tables or code blocks. The original is authoritative.
Published by arXiv cs.RO
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗