Research

Paper

AI LLM February 24, 2026

Efficient and Explainable End-to-End Autonomous Driving via Masked Vision-Language-Action Diffusion

Authors

Jiaru Zhang, Manav Gagvani, Can Cui, Juntong Peng, Ruqi Zhang, Ziran Wang

Abstract

Large Language Models (LLMs) and Vision-Language Models (VLMs) have emerged as promising candidates for end-to-end autonomous driving. However, these models typically face challenges in inference latency, action precision, and explainability. Existing autoregressive approaches struggle with slow token-by-token generation, while prior diffusion-based planners often rely on verbose, general-purpose language tokens that lack explicit geometric structure. In this work, we propose Masked Vision-Language-Action Diffusion for Autonomous Driving (MVLAD-AD), a novel framework designed to bridge the gap between efficient planning and semantic explainability via a masked vision-language-action diffusion model. Unlike methods that force actions into the language space, we introduce a discrete action tokenization strategy that constructs a compact codebook of kinematically feasible waypoints from real-world driving distributions. Moreover, we propose geometry-aware embedding learning to ensure that embeddings in the latent space approximate physical geometric metrics. Finally, an action-priority decoding strategy is introduced to prioritize trajectory generation. Extensive experiments on nuScenes and derived benchmarks demonstrate that MVLAD-AD achieves superior efficiency and outperforms state-of-the-art autoregressive and diffusion baselines in planning precision, while providing high-fidelity and explainable reasoning.

Metadata

arXiv ID: 2602.20577
Provider: ARXIV
Primary Category: cs.CV
Published: 2026-02-24
Fetched: 2026-02-25 06:05

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.20577v1</id>\n    <title>Efficient and Explainable End-to-End Autonomous Driving via Masked Vision-Language-Action Diffusion</title>\n    <updated>2026-02-24T05:59:10Z</updated>\n    <link href='https://arxiv.org/abs/2602.20577v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.20577v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Large Language Models (LLMs) and Vision-Language Models (VLMs) have emerged as promising candidates for end-to-end autonomous driving. However, these models typically face challenges in inference latency, action precision, and explainability. Existing autoregressive approaches struggle with slow token-by-token generation, while prior diffusion-based planners often rely on verbose, general-purpose language tokens that lack explicit geometric structure. In this work, we propose Masked Vision-Language-Action Diffusion for Autonomous Driving (MVLAD-AD), a novel framework designed to bridge the gap between efficient planning and semantic explainability via a masked vision-language-action diffusion model. Unlike methods that force actions into the language space, we introduce a discrete action tokenization strategy that constructs a compact codebook of kinematically feasible waypoints from real-world driving distributions. Moreover, we propose geometry-aware embedding learning to ensure that embeddings in the latent space approximate physical geometric metrics. Finally, an action-priority decoding strategy is introduced to prioritize trajectory generation. Extensive experiments on nuScenes and derived benchmarks demonstrate that MVLAD-AD achieves superior efficiency and outperforms state-of-the-art autoregressive and diffusion baselines in planning precision, while providing high-fidelity and explainable reasoning.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CV'/>\n    <published>2026-02-24T05:59:10Z</published>\n    <arxiv:primary_category term='cs.CV'/>\n    <author>\n      <name>Jiaru Zhang</name>\n    </author>\n    <author>\n      <name>Manav Gagvani</name>\n    </author>\n    <author>\n      <name>Can Cui</name>\n    </author>\n    <author>\n      <name>Juntong Peng</name>\n    </author>\n    <author>\n      <name>Ruqi Zhang</name>\n    </author>\n    <author>\n      <name>Ziran Wang</name>\n    </author>\n  </entry>"
}