HiFlow: Tokenization-Free Scale-Wise Autoregressive Policy Learning via Flow Matching

TL;DR

HiFlow introduces a tokenization-free, hierarchical autoregressive policy that directly models continuous robot actions, improving efficiency and performance over existing tokenization-based methods.

Contribution

It proposes a novel hierarchical flow policy (HiFlow) that operates directly on continuous actions, eliminating the need for tokenization and multi-stage training.

Findings

HiFlow outperforms diffusion-based policies in experiments.

HiFlow achieves better results than tokenization-based autoregressive policies.

The model trains end-to-end in a single stage.

Abstract

Coarse-to-fine autoregressive modeling has recently shown strong promise for visuomotor policy learning, combining the inference efficiency of autoregressive methods with the global trajectory coherence of diffusion-based policies. However, existing approaches rely on discrete action tokenizers that map continuous action sequences to codebook indices, a design inherited from image generation where learned compression is necessary for high-dimensional pixel data. We observe that robot actions are inherently low-dimensional continuous vectors, for which such tokenization introduces unnecessary quantization error and a multi-stage training pipeline. In this work, we propose Hierarchical Flow Policy (HiFlow), a tokenization-free coarse-to-fine autoregressive policy that operates directly on raw continuous actions. HiFlow constructs multi-scale continuous action targets from each action chunk via simple temporal pooling. Specifically, it averages contiguous action windows to produce coarse summaries that are refined at finer temporal resolutions. The entire model is trained end-to-end in a single stage, eliminating the need for a separate tokenizer. Experiments on MimicGen, RoboTwin 2.0, and real-world environments demonstrate that HiFlow consistently outperforms existing methods including diffusion-based and tokenization-based autoregressive policies.