DriveFine: Refining-Augmented Masked Diffusion VLA for Precise and Robust Driving

TL;DR

DriveFine introduces a novel masked diffusion model with a plug-and-play expert refinement mechanism for autonomous driving, improving robustness and flexibility over existing VLA models.

Contribution

The paper presents DriveFine, a diffusion-based VLA model with a decoupled expert refinement component and a hybrid reinforcement learning strategy, enhancing performance and generalization.

Findings

Outperforms existing models on NAVSIM v1, v2, and Navhard benchmarks.

Demonstrates strong robustness and efficacy in autonomous driving tasks.

Shows effective self-correction and flexible decoding capabilities.

Abstract

Vision-Language-Action (VLA) models for autonomous driving increasingly adopt generative planners trained with imitation learning followed by reinforcement learning. Diffusion-based planners suffer from modality alignment difficulties, low training efficiency, and limited generalization. Token-based planners are plagued by cumulative causal errors and irreversible decoding. In summary, the two dominant paradigms exhibit complementary strengths and weaknesses. In this paper, we propose DriveFine, a masked diffusion VLA model that combines flexible decoding with self-correction capabilities. In particular, we design a novel plug-and-play block-MoE, which seamlessly injects a refinement expert on top of the generation expert. By enabling explicit expert selection during inference and gradient blocking during training, the two experts are fully decoupled, preserving the foundational capabilities and generic patterns of the pretrained weights, which highlights the flexibility and extensibility of the block-MoE design. Furthermore, we design a hybrid reinforcement learning strategy that encourages effective exploration of refinement expert while maintaining training stability. Extensive experiments on NAVSIM v1, v2, and Navhard benchmarks demonstrate that DriveFine exhibits strong efficacy and robustness. The code will be released at https://github.com/MSunDYY/DriveFine.