DFM-VLA: Iterative Action Refinement for Robot Manipulation via Discrete Flow Matching

2026-03-27

AI Summary

This work addresses the limitation of existing vision-language-action (VLA) models, which struggle to ensure execution accuracy due to their inability to revise early-generated action tokens. To overcome this, we propose DFM-VLA, the first VLA framework incorporating discrete flow matching. By modeling token-level probability velocity fields, DFM-VLA dynamically refines the entire action sequence over multiple iterations. The method employs a two-stage decoding strategy to ensure stable convergence and investigates two formulations for velocity field modeling: an auxiliary velocity head and action embedding guidance. Evaluated on the CALVIN and LIBERO benchmarks, DFM-VLA achieves an average success length of 4.44 and an average success rate of 95.7%, significantly outperforming both autoregressive and diffusion-based baselines while maintaining efficient inference.

Abstract

Vision--Language--Action (VLA) models that encode actions using a discrete tokenization scheme are increasingly adopted for robotic manipulation, but existing decoding paradigms remain fundamentally limited. Whether actions are decoded sequentially by autoregressive VLAs or in parallel by discrete diffusion VLAs, once a token is generated, it is typically fixed and cannot be revised in subsequent iterations, so early token errors cannot be effectively corrected later. We propose DFM-VLA, a discrete flow matching VLA for iterative refinement of action tokens. DFM-VLA~models a token-level probability velocity field that dynamically updates the full action sequence across refinement iterations. We investigate two ways to construct the velocity field: an auxiliary velocity-head formulation and an action-embedding-guided formulation. Our framework further adopts a two-stage decoding strategy with an iterative refinement stage followed by deterministic validation for stable convergence. Extensive experiments on CALVIN, LIBERO, and real-world manipulation tasks show that DFM-VLA consistently outperforms strong autoregressive, discrete diffusion, and continuous diffusion baselines in manipulation performance while retaining high inference efficiency. In particular, DFM-VLA achieves an average success length of 4.44 on CALVIN and an average success rate of 95.7\% on LIBERO, highlighting the value of action refinement via discrete flow matching for robotic manipulation. Our project is available \url{https://chris1220313648.github.io/DFM-VLA/}

Problem

• Vision-Language-Action
• discrete tokenization
• action refinement
• robotic manipulation
• iterative decoding

Innovation

• Discrete Flow Matching
• Vision-Language-Action Models
• Iterative Action Refinement
• Token-level Velocity Field
• Robotic Manipulation