aiqbench analysis
This research introduces TOAST, a stochastic tokenization method that leverages representational redundancy in discrete action sequences to enhance policy learning for autoregressive robot models. By sampling alternative tokenizations during training instead of using deterministic mappings, the approach significantly boosts success rates with limited demonstration data, particularly on real-robot manipulation tasks. The findings highlight a novel strategy for improving data efficiency in next-token prediction frameworks without requiring additional training samples.
TOAST: Stochastic Robot Action Tokenization for Autoregressive Vision-Language-Action Models
arXiv:2610.00899v1 Announce Type: cross Abstract: Autoregressive Vision-Language-Action models often represent continuous robot actions as discrete token sequences, enabling action prediction with standard next-token objectives. FAST has substantially improved this representation by compactly encoding action containing diverse temporal frequencies into relatively few tokens. However, while such compression reduces the number of action tokens required for autoregressive prediction, it does not necessarily improve the efficiency of policy learning from limited demonstrations. In particular, FAST typically assigns a single deterministic tokenization to each quantized action sequence, although multiple token sequences can represent and decode to the same robot motion. We investigate whether exploiting this representational redundancy can improve policy learning. In this paper, we propose TOkenization of Action sequences with STochastic sampling (TOAST), a stochastic action tokenization method that samples alternative tokenizations of the same quantized action sequence during policy training. This diversifies the discrete supervision while preserving the underlying robot action and requires no additional demonstrations. Experiments on LIBERO show that TOAST consistently improves over its deterministic counterpart, with the improvement increasing as training data decreases, achieving a 6.8 point gain in success rate when only 1/16 of training data is available. Across four real-robot manipulation tasks, TOAST further improves mean success rate by 15.8 points over the deterministic counterpart. These results demonstrate the effectiveness of stochastic action tokenization for autoregressive robot policy learning, particularly when training data are limited.
Abstract:Autoregressive Vision-Language-Action models often represent continuous robot actions as discrete token sequences, enabling action prediction with standard next-token objectives. FAST has substantially improved this representation by compactly encoding action containing diverse temporal frequencies into relatively few tokens. However, while such compression reduces the number of action tokens required for autoregressive prediction, it does not necessarily improve the efficiency of policy learning from limited demonstrations. In particular, FAST typically assigns a single deterministic tokenization to each quantized action sequence, although multiple token sequences can represent and decode to the same robot motion. We investigate whether exploiting this representational redundancy can improve policy learning. In this paper, we propose TOkenization of Action sequences with STochastic sampling (TOAST), a stochastic action tokenization method that samples alternative tokenizations of the same quantized action sequence during policy training. This diversifies the discrete supervision while preserving the underlying robot action and requires no additional demonstrations. Experiments on LIBERO show that TOAST consistently improves over its deterministic counterpart, with the improvement increasing as training data decreases, achieving a 6.8 point gain in success rate when only 1/16 of training data is available. Across four real-robot manipulation tasks, TOAST further improves mean success rate by 15.8 points over the deterministic counterpart. These results demonstrate the effectiveness of stochastic action tokenization for autoregressive robot policy learning, particularly when training data are limited.
Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2610.00899 [cs.RO]
(or arXiv:2610.00899v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.00899 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 · 322 words
Related coverage
Steering Autoregressive Vision-Language-Action Policies via Action Token Intervention
arXiv · Jun 16, 2026
Runway’s WorldPrompt and the Engineering of Real-Time Worlds
Latent Space · Sep 25, 2026
Fast Plans, Faithful Actions: Closing the Planning-Execution Gap in Hierarchical Vision-Language-Action Models
arXiv · Sep 28, 2026
SUAVE: Unified Video-Action Models via Masked Diffusion
arXiv · Oct 6, 2026