Research map / Vision-language and multimodal / Robot action and navigation
DynVLA: Learning World Dynamics for Action Reasoning in Autonomous Driving
Shuyao Shang, Bing Zhan, Yunfei Yan, Yuqi Wang, Yingyan Li, Yasong An, Xiaoman Wang, Jierui Liu, Lu Hou, Lue Fan, Zhaoxiang Zhang, Tieniu Tan
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 |
|---|---|---|---|
| use Dynamics CoT instead of textual scene description CoT (ablation) | DynVLA (EMU3) · NAVSIM | PDMS: 85.3 -> 87.2 | Improved |
| add Dynamics CoT SFT (ablation) | EMU3 baseline · NAVSIM | PDMS: 85.6 -> 87.2 | Improved |
| add RFT on top of Dynamics CoT SFT (ablation) | DynVLA (EMU3) · NAVSIM | PDMS: 87.2 -> 91.7 | Improved |
| apply RFT to baseline without CoT (ablation) | EMU3 baseline · NAVSIM | PDMS: 85.6 -> 88.7 | Improved |
| use Dynamics CoT with Qwen2.5-VL base instead of no CoT (ablation) | Qwen2.5-VL · NAVSIM | PDMS: 85.3 -> 86.6 (SFT), 91.0 (SFT+RFT) | Improved |
| decouple ego-centric and environment-centric dynamics with ego action regularization (ablation) | Dynamics Tokenizer · NAVSIM | PDMS: 86.2 -> 87.2 | Improved |
| decoupling dynamics to avoid codebook collapse (ablation) | Dynamics Tokenizer · VQ code usage during tokenizer training | activated VQ codes | Improved |
| add BEV branch (cross-view consistency) alongside image branch (ablation) | Dynamics Tokenizer · NAVSIM | PDMS: 86.7 -> 87.2 | Improved |
| add image branch alongside BEV branch (ablation) | Dynamics Tokenizer · NAVSIM | PDMS: 86.2 -> 87.2 | Improved |
| extend prediction horizon from K=2 to K=3 (ablation) | Dynamics CoT · NAVSIM | PDMS: 87.2 -> 86.7 | Hurt |
| extend prediction horizon from K=1 to K=2 (ablation) | Dynamics CoT · NAVSIM | PDMS: 86.5 -> 87.2 | Improved |
| increase dynamics tokens from 8 to 16 (ablation) | Dynamics Tokenizer · NAVSIM | PDMS: 87.2 -> 86.5 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 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
- DriveVLA-W0 — DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving (ICLR 2026) · better on 1 of 1 shared result
- SimLingo — SimLingo: Vision-Only Closed-Loop Autonomous Driving with Language-Action Alignment (CVPR 2025) · 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
- OneVL: One-Step Latent Reasoning and Planning with Vision-Language Explanation (NeurIPS 2026 · similarity 0.89)
- FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving (NeurIPS 2025 · similarity 0.89)
- AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (NeurIPS 2025 · similarity 0.89)
- LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving (NeurIPS 2026 · similarity 0.88)
- Dyn-O: Building Structured World Models with Object-Centric Representations (NeurIPS 2025 · similarity 0.87)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · similarity 0.87)
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.
