DriveMA: Rethinking Language Interfaces in Driving VLAs with One-Step Meta-Actions

• URL: http://arxiv.org/abs/2605.21273v2
• Date: Thu, 21 May 2026 07:58:08 GMT
• Title: DriveMA: Rethinking Language Interfaces in Driving VLAs with One-Step Meta-Actions
• Abstract summary: In this paper, we show that concise one-step meta-actions are a simple yet effective alternative to verbose reasoning.<n>We propose DriveMA, which combines action-centric supervised training with a turn-level credit-assignment reinforcement learning framework.<n>DriveMA achieves a new state of the art on the End-to-End Driving Challenge with a 2B model, reaching a Rater Feedback Score (RFS) of 8.060, while its 4B version further improves the state of the art to 8.079.
• Score: 31.04597401352966
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Driving Vision-Language-Action Models (Driving VLAs) commonly introduce natural-language reasoning as an intermediate interface for end-to-end planning, but reasoning-centric interfaces face three practical bottlenecks: obtaining high-quality reasoning annotations is difficult, generating and understanding long reasoning chains is challenging for compact models, and inference latency is substantially increased. In this paper, we rethink the design of language interfaces in Driving VLAs and show that concise one-step meta-actions are a simple yet effective alternative to verbose reasoning. Meta-actions provide semantic decision grounding while remaining low-entropy, and being automatically derivable from expert trajectories, enabling scalable supervision and reliable trajectory conditioning. Building on this interface, we propose DriveMA, which combines action-centric supervised training with a turn-level credit-assignment reinforcement learning framework that jointly optimizes meta-action correctness, trajectory quality, and trajectory--meta-action consistency. Experiments show that DriveMA already achieves a new state of the art on the Waymo End-to-End Driving Challenge with a 2B model, reaching a Rater Feedback Score (RFS) of 8.060, while its 4B version further improves the state of the art to 8.079; DriveMA also obtains competitive performance on NAVSIM. Ablations demonstrate that one-step meta-actions offer a better practical trade-off between expressiveness, predictability, and inference efficiency than natural-language reasoning or finer-grained action sequences. Code, data, and models will be released to facilitate future research.

Related papers

• DriveMA: Driving Vision-Language-Action Models with verifiable Meta-Actions [31.04597401352966]
  We propose DriveMA, a Driving VLA framework built on verifiable meta-actions.<n>DriveMA exploits this verifiability with action-centric supervised training and a data-efficient turn-level credit assignment reinforcement learning framework.<n>We show that even a simple meta-action interface can achieve state-of-the-art planning when made verifiable and optimized for language-action alignment.
  arXiv  Detail & Related papers  (2026-05-29T13:03:16Z)
• AnE: Pushing the Reasoning Frontier of Multimodal LLMs via Anchor Evolution [61.593935260052795]
  Post-training viaSupervised Fine-Tuning (SFT) and Reinforcement Learning (RL) is crucial for enhancing reasoning in Multimodal Large Language Models (MLLMs)<n>Existing paradigms often reach a performance bottleneck due to the limitations of static data.<n>We propose Anchor Evolution (AnE), a new paradigm that integrates truth-anchored data curation and model evolution.
  arXiv  Detail & Related papers  (2026-05-25T08:26:34Z)
• Unifying Language-Action Understanding and Generation for Autonomous Driving [25.23561391638388]
  Vision-Language-Action (VLA) models are emerging as a promising paradigm for end-to-end autonomous driving.<n>Existing methods suffer from two critical limitations: a persistent misalignment between language instructions and action outputs, and the inherent inefficiency of typical auto-regressive action generation.<n>We introduce LinkVLA, a novel architecture that directly addresses these challenges to enhance both alignment and efficiency.
  arXiv  Detail & Related papers  (2026-03-02T04:41:10Z)
• Efficient and Explainable End-to-End Autonomous Driving via Masked Vision-Language-Action Diffusion [23.834662472392694]
  Masked Vision-Language-Action Diffusion for Autonomous Driving (MVLAD-AD) is a novel framework designed to bridge the gap between efficient planning and semantic explainability.<n>We introduce a discrete action tokenization strategy that constructs a compact codebook of kinematically feasible waypoints from real-world driving distributions.<n>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.
  arXiv  Detail & Related papers  (2026-02-24T05:59:10Z)
• Vision-Language-Action Models for Autonomous Driving: Past, Present, and Future [125.92052530850425]
  Vision-Language-Action (VLA) frameworks integrate perception with language-grounded decision making.<n>VLA frameworks offer a pathway toward more interpretable, generalizable, and human-aligned driving policies.<n>This work aims to establish a coherent foundation for advancing human-compatible autonomous driving systems.
  arXiv  Detail & Related papers  (2025-12-18T16:57:44Z)
• Latent Chain-of-Thought World Modeling for End-to-End Driving [45.726304769312414]
  We present Latent-CoT-Drive (LCDrive), a model that expresses CoT in a latent language.<n>Our approach unifies CoT reasoning and decision making by representing both in an action-aligned latent space.<n>On a large-scale end-to-end driving benchmark, LCDrive achieves faster inference, better trajectory quality, and larger improvements from interactive reinforcement learning.
  arXiv  Detail & Related papers  (2025-12-11T02:22:07Z)
• Discrete Diffusion for Reflective Vision-Language-Action Models in Autonomous Driving [55.13109926181247]
  We introduce ReflectDrive, a learning-based framework that integrates a reflection mechanism for safe trajectory generation via discrete diffusion.<n>Central to our approach is a safety-aware reflection mechanism that performs iterative self-correction without gradient.<n>Our method begins with goal-conditioned trajectory generation to model multi-modal driving behaviors.
  arXiv  Detail & Related papers  (2025-09-24T13:35:15Z)
• Sycophancy Mitigation Through Reinforcement Learning with Uncertainty-Aware Adaptive Reasoning Trajectories [58.988535279557546]
  We introduce textbf sycophancy Mitigation through Adaptive Reasoning Trajectories.<n>We show that SMART significantly reduces sycophantic behavior while preserving strong performance on out-of-distribution inputs.
  arXiv  Detail & Related papers  (2025-09-20T17:09:14Z)
• AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning [37.176428069948535]
  Vision-Language-Action (VLA) models have shown promise for end-to-end autonomous driving.<n>Current VLA models struggle with physically infeasible action outputs, complex model structures, or unnecessarily long reasoning.<n>We propose AutoVLA, a novel VLA model that unifies reasoning and action generation within a single autoregressive generation model.
  arXiv  Detail & Related papers  (2025-06-16T17:58:50Z)
• ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Driving [49.07731497951963]
  ReCogDrive is a novel Reinforced Cognitive framework for end-to-end autonomous driving.<n>We introduce a hierarchical data pipeline that mimics the sequential cognitive process of human drivers.<n>We then address the language-action mismatch by injecting the VLM's learned driving priors into a diffusion planner.
  arXiv  Detail & Related papers  (2025-06-09T03:14:04Z)
• DiFSD: Ego-Centric Fully Sparse Paradigm with Uncertainty Denoising and Iterative Refinement for Efficient End-to-End Self-Driving [55.53171248839489]
  We propose an ego-centric fully sparse paradigm, named DiFSD, for end-to-end self-driving.<n>Specifically, DiFSD mainly consists of sparse perception, hierarchical interaction and iterative motion planner.<n>Experiments conducted on nuScenes and Bench2Drive datasets demonstrate the superior planning performance and great efficiency of DiFSD.
  arXiv  Detail & Related papers  (2024-09-15T15:55:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.