aiqbench analysis
This paper introduces Residual Action Tokenization (ReAT), a method that enhances autoregressive robot policies by using decoded reconstructions as feedback to refine subsequent action tokens. By closing the loop between encoding and decoding, ReAT enables more efficient self-corrective learning, allowing shorter token sequences to outperform existing baselines across diverse manipulation tasks in both simulation and real-world settings.
ReAT: Residual Action Tokenization for Autoregressive Robot Policies
arXiv:2610.13093v1 Announce Type: new Abstract: Autoregressive robot policies rely on action tokenizers to convert continuous action chunks into compact sequences of discrete prediction targets. Existing tokenizers can reconstruct action chunks from discrete codes, but often leave implicit what each individual token contributes to the reconstructed action. We introduce Residual Action Tokenization (ReAT), which gives each token an explicit role in refining the current action reconstruction. It uses the decoded reconstruction as feedback to construct the next token from the remaining action error. This closes the loop between encoding and decoding, making each new token responsive to the errors left by the actual decoded prefix. Across 19 manipulation tasks spanning three simulation benchmarks and real-world settings, 8-token ReAT policies outperform the evaluated baselines in task success; on LIBERO, even 4-token ReAT surpasses the evaluated 8-token action-tokenization baselines at half the token budget. These results suggest a path toward action tokenization that supports self-corrective policy learning, helping autoregressive policies learn to progressively refine their own action plans.
Abstract:Autoregressive robot policies rely on action tokenizers to convert continuous action chunks into compact sequences of discrete prediction targets. Existing tokenizers can reconstruct action chunks from discrete codes, but often leave implicit what each individual token contributes to the reconstructed action. We introduce Residual Action Tokenization (ReAT), which gives each token an explicit role in refining the current action reconstruction. It uses the decoded reconstruction as feedback to construct the next token from the remaining action error. This closes the loop between encoding and decoding, making each new token responsive to the errors left by the actual decoded prefix. Across 19 manipulation tasks spanning three simulation benchmarks and real-world settings, 8-token ReAT policies outperform the evaluated baselines in task success; on LIBERO, even 4-token ReAT surpasses the evaluated 8-token action-tokenization baselines at half the token budget. These results suggest a path toward action tokenization that supports self-corrective policy learning, helping autoregressive policies learn to progressively refine their own action plans.
Robotics (cs.RO)
arXiv:2610.13093 [cs.RO]
(or arXiv:2610.13093v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.13093 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 253 words
Related coverage
Ordered Action Tokens for Visuomotor Policy Learning
arXiv · Jul 28, 2026
NAC: Neural Action Codec for Vision-Language-Action Models
arXiv · Sep 28, 2026
Imagine-RL: Residual-Confidence-Guided Cross-Attention for World-Model-Augmented VLA Reinforcement Learning
arXiv · Sep 22, 2026
TOAST: Stochastic Robot Action Tokenization for Autoregressive Vision-Language-Action Models
arXiv · Oct 2, 2026