Research map / Vision-language and multimodal / Robot action and navigation
DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving
Yingyan Li, Shuyao Shang, Weisong Liu, Bing Zhan, Haochen Wang, Yuqi Wang, Yuntao Chen, Xiaoman Wang, AnYasong, Chufeng Tang, Lu Hou, Lue Fan and 1 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add action grounding (interleaved vision-action 6VA) to vision-only world-model pretraining (ablation) | VLA-W0-VQ · NAVSIM | PDMS: 84.1 -> 85.6 | Improved |
| lengthen pretraining sequence from VA to 6VA (ablation) | VLA-W0-VQ · NAVSIM | PDMS: 83.3 -> 85.6 | Improved |
| lengthen pretraining sequence from 2VA to 6VA (ablation) | VLA-W0-VQ · NAVSIM | PDMS: 84.2 -> 85.6 | Improved |
| add NuPlan world-model pretraining before NAVSIM finetuning (ablation) | VLA-W0-VQ · NAVSIM | PDMS: 80.7 -> 85.6 | Improved |
| add NuPlan pretraining to action-only VLA baseline | VLA-VQ · NAVSIM | PDMS: 68.7 -> 62.2 | Hurt |
| add NuPlan pretraining to TransFuser-7B | TransFuser-7B · NAVSIM | PDMS: 77.9 -> 71.6 | Hurt |
| add world modeling to VQ VLA at 70M frames | VLA (VQ) · In-house dataset (70M frames) | ADE (m): 1.4829 -> 1.0563 | Improved |
| add world modeling to ViT VLA at 70M frames | VLA (ViT) · In-house dataset (70M frames) | Collision (%): 0.0359 -> 0.0302 | Improved |
| add world modeling to ViT VLA at 700k frames | VLA (ViT) · In-house dataset (700k frames) | Collision (%): 0.0462 -> 0.0513 | Hurt |
| add world modeling to VQ VLA at 700k frames | VLA (VQ) · In-house dataset (700k frames) | ADE (m): 1.5424 -> 1.5985 | Hurt |
| swap query-based action expert for autoregressive expert at 70M frames | VLA · In-house dataset (70M frames) | Collision (%): 0.0453 -> 0.0295 | Improved |
| swap query-based action expert for autoregressive expert on NAVSIM | VLA · NAVSIM (103k frames) | PDMS: 88.4 -> 85.3 | Hurt |
Later papers that compared with this method
8 later accepted papers report DriveVLA-W0 as a baseline in their results tables; 8 of them report a better number on most of the results they share with it.
- DriveFine: Refining-Augmented Masked Diffusion VLA for Accurate and Robust Driving (ECCV 2026 · better on 3 of 3 shared results)
- DriveWorld-VLA: Unified Latent-Space World Modeling with Vision–Language–Action for Autonomous Driving (ICML 2026 · better on 2 of 2 shared results)
- DriveDreamer-Policy: A Geometry-Grounded World-Action Model for Unified Generation and Planning (NeurIPS 2026 · better on 2 of 2 shared results)
- ExploreVLA: Dense World Modeling and Exploration for End-to-End Autonomous Driving (ECCV 2026 · better on 2 of 2 shared results)
- DynVLA: Learning World Dynamics for Action Reasoning in Autonomous Driving (ICML 2026 · better on 1 of 1 shared result)
- Intend, Reflect, Refine: An Adaptive Multimodal Reflection Framework for Autonomous Driving (NeurIPS 2026 · better on 1 of 1 shared result)
- Think Densely, Act Sparsely: Latent Expert Cognitive Chains for Vision-Language-Action Autonomous Driving (NeurIPS 2026 · better on 1 of 1 shared result)
- ChainFlow-VLA: Causal Flow Planning with Vision-Language Models (NeurIPS 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ReCogDrive — ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Driving (ICLR 2026) · better on 1 of 1 shared result
- AutoVLA — AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (NeurIPS 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Impromptu VLA: Open Weights and Open Data for Driving Vision-Language-Action Models (NeurIPS 2025 · similarity 0.92)
- Less Language, More Latents: Annotation-Efficient VLAs for Driving (NeurIPS 2026 · similarity 0.91)
- FIVE-VLA: Fast and EffectIVE Closed-Loop Autonomous Driving with Recurrent Action Memory (NeurIPS 2026 · similarity 0.91)
- AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (NeurIPS 2025 · similarity 0.90)
- LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving (NeurIPS 2026 · similarity 0.90)
- CoC-VLA: Delving into Adversarial Domain Transfer for Explainable Autonomous Driving via Chain-of-Causality Visual-Language-Action Model (NeurIPS 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
