DriveMA: Driving Vision-Language-Action Models with verifiable Meta-Actions
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913172865155072 |
|---|---|
| author | Zheng, Weicheng Huang, Yixin Sun, Qiao Li, Derun Zhao, Hang |
| author_facet | Zheng, Weicheng Huang, Yixin Sun, Qiao Li, Derun Zhao, Hang |
| contents | Driving Vision-Language-Action Models (Driving VLAs) aim to use language to improve end-to-end planning, but the language-action gap limits this promise. We propose DriveMA, a Driving VLA framework built on verifiable meta-actions, which summarize future ego motion into compact language-domain intentions and can be constructed from expert trajectories with a trajectory-grounded annotation pipeline and can be verified against generated trajectories through rule-based projection. DriveMA exploits this verifiability with action-centric supervised training and a data-efficient turn-level credit assignment reinforcement learning framework, explicitly aligning high-level decisions with low-level trajectory planning through dense rewards and precise credit assignment. DriveMA sets a new state of the art on the Waymo Open Dataset Vision-based E2E Driving, achieving a Rater Feedback Score of 8.060 with a 2B model and further improving it to 8.079 with a 4B model; it also obtains competitive closed-loop planning performance on NAVSIM. These results show that even a simple meta-action interface can achieve state-of-the-art planning when made verifiable and optimized for language-action alignment. Code, data, and models will be released to facilitate future research. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_31271 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | DriveMA: Driving Vision-Language-Action Models with verifiable Meta-Actions Zheng, Weicheng Huang, Yixin Sun, Qiao Li, Derun Zhao, Hang Computer Vision and Pattern Recognition Driving Vision-Language-Action Models (Driving VLAs) aim to use language to improve end-to-end planning, but the language-action gap limits this promise. We propose DriveMA, a Driving VLA framework built on verifiable meta-actions, which summarize future ego motion into compact language-domain intentions and can be constructed from expert trajectories with a trajectory-grounded annotation pipeline and can be verified against generated trajectories through rule-based projection. DriveMA exploits this verifiability with action-centric supervised training and a data-efficient turn-level credit assignment reinforcement learning framework, explicitly aligning high-level decisions with low-level trajectory planning through dense rewards and precise credit assignment. DriveMA sets a new state of the art on the Waymo Open Dataset Vision-based E2E Driving, achieving a Rater Feedback Score of 8.060 with a 2B model and further improving it to 8.079 with a 4B model; it also obtains competitive closed-loop planning performance on NAVSIM. These results show that even a simple meta-action interface can achieve state-of-the-art planning when made verifiable and optimized for language-action alignment. Code, data, and models will be released to facilitate future research. |
| title | DriveMA: Driving Vision-Language-Action Models with verifiable Meta-Actions |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2605.31271 |