MARS Policy: Multimodality Only When It Matters

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Jia, Jindou, An, Tuo, Hu, Yuxuan, Li, Gen, Li, Jingliang, Hou, Bohan, Chen, Xiangyu, Bai, Jiaqi, Lyu, Bofan, Yang, Jianfei
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866916059418722304
author Jia, Jindou
An, Tuo
Hu, Yuxuan
Li, Gen
Li, Jingliang
Hou, Bohan
Chen, Xiangyu
Bai, Jiaqi
Lyu, Bofan
Yang, Jianfei
author_facet Jia, Jindou
An, Tuo
Hu, Yuxuan
Li, Gen
Li, Jingliang
Hou, Bohan
Chen, Xiangyu
Bai, Jiaqi
Lyu, Bofan
Yang, Jianfei
contents Imitation learning has become a cornerstone for solving complex robotic manipulation tasks. In particular, multimodality, which enables robots to capture diverse yet valid behavioral patterns, has driven the rapid emergence of generative policies as a dominant paradigm in robot learning. However, achieving such multimodality typically relies on stochastic noise initialization and iterative denoising procedures, resulting in substantial training complexity and low inference efficiency. Meanwhile, not all phases of a robotic task inherently require behavioral diversity. Motivated by this insight, we propose the Modality-Adaptive Robot Sampling (MARS) policy, which adaptively invokes tailored stochasticity only when it is truly beneficial, while reverting to an efficient deterministic learning during single-modal phases. In other words, the proper amount of noise is injected only at the proper time. By selectively activating multimodal generation, MARS policy bridges the gap between the multimodal capability of generative policies and the superior training and inference efficiency of deterministic models. Empirical studies across 8 simulated and 4 real-world tasks demonstrate that MARS exhibits robust multimodal expressivity and high efficiency, with a 16.67% success rate improvement and an 83.20% inference latency reduction in real-world tests. Counterintuitively, MARS also outpaces deterministic policies in training efficiency on near-deterministic tasks by more effectively modeling nuanced action diversity.
format Preprint
id arxiv_https___arxiv_org_abs_2605_29766
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MARS Policy: Multimodality Only When It Matters
Jia, Jindou
An, Tuo
Hu, Yuxuan
Li, Gen
Li, Jingliang
Hou, Bohan
Chen, Xiangyu
Bai, Jiaqi
Lyu, Bofan
Yang, Jianfei
Robotics
Imitation learning has become a cornerstone for solving complex robotic manipulation tasks. In particular, multimodality, which enables robots to capture diverse yet valid behavioral patterns, has driven the rapid emergence of generative policies as a dominant paradigm in robot learning. However, achieving such multimodality typically relies on stochastic noise initialization and iterative denoising procedures, resulting in substantial training complexity and low inference efficiency. Meanwhile, not all phases of a robotic task inherently require behavioral diversity. Motivated by this insight, we propose the Modality-Adaptive Robot Sampling (MARS) policy, which adaptively invokes tailored stochasticity only when it is truly beneficial, while reverting to an efficient deterministic learning during single-modal phases. In other words, the proper amount of noise is injected only at the proper time. By selectively activating multimodal generation, MARS policy bridges the gap between the multimodal capability of generative policies and the superior training and inference efficiency of deterministic models. Empirical studies across 8 simulated and 4 real-world tasks demonstrate that MARS exhibits robust multimodal expressivity and high efficiency, with a 16.67% success rate improvement and an 83.20% inference latency reduction in real-world tests. Counterintuitively, MARS also outpaces deterministic policies in training efficiency on near-deterministic tasks by more effectively modeling nuanced action diversity.
title MARS Policy: Multimodality Only When It Matters
topic Robotics
url https://arxiv.org/abs/2605.29766