aiqbench analysis
The paper tests whether flow-matching vision-language-action robot policies need multi-step ODE integration at inference, or whether a single step is enough. Across six models, one-step generation largely matches multi-step sampling once models are well trained and have enough capacity. For practitioners, this suggests lower-latency robot control may not require the distillation or auxiliary objectives that recent single-shot methods add, though the result depends on model scale and training quality and needs independent replication.
When One Step Is Enough: An Empirical Study of Euler Integration in Flow-Matching Policies
From arXiv
Full textarXiv:2610.30735v1 Announce Type: new Abstract: Flow matching is now the dominant approach for vision-language action (VLA) robot policies, thanks to its expressiveness and simple training objective. However, inference for standard flow matching requires iteratively solving an ODE, limiting the control frequency at deployment. While several works have proposed novel methods to allow for single-shot action generation, we ask whether such additional objectives are actually necessary. We find that across six flow matching VLAs, one-step action generation largely matches multi-step sampling at scale: multi-step integration confers little to no consistent advantage at inference.
We empirically find that this holds as long as models are sufficiently trained and have enough capacity. Practically, the implication is that for large VLAs, the single-step policy comes essentially for free from the multi-step model, enabling low-latency deployment without retraining, distillation, or auxiliary objectives.
Abstract:Flow matching is now the dominant approach for vision-language action (VLA) robot policies, thanks to its expressiveness and simple training objective. However, inference for standard flow matching requires iteratively solving an ODE, limiting the control frequency at deployment. While several works have proposed novel methods to allow for single-shot action generation, we ask whether such additional objectives are actually necessary. We find that across six flow matching VLAs, one-step action generation largely matches multi-step sampling at scale: multi-step integration confers little to no consistent advantage at inference.
We empirically find that this holds as long as models are sufficiently trained and have enough capacity. Practically, the implication is that for large VLAs, the single-step policy comes essentially for free from the multi-step model, enabling low-latency deployment without retraining, distillation, or auxiliary objectives.
Robotics (cs.RO)
arXiv:2610.30735 [cs.RO]
(or arXiv:2610.30735v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.30735 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 · 225 words