Research map / Vision-language and multimodal / Robot action and navigation
Unifying Language-Action Understanding and Generation for Autonomous Driving
Xinyang Wang, Qian Liu, Wenjie Ding, Zhao Yang, Wei Li, Chang Liu, Bailin Li, Kun Zhan, Xianpeng Lang, Wei Chen
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 unified action tokenization (log coordinates, shared codebook) (ablation) | LinkVLA · Bench2Drive | Driving Score: 85.07 -> 89.57 | Improved |
| add coarse-to-fine generation after tokenization (ablation) | LinkVLA · Bench2Drive | Success Rate (%): 73.18 -> 72.27 | No clear effect |
| add action understanding alignment objective (ablation) | LinkVLA · Bench2Drive | Driving Score: 89.85 -> 91.01 | Improved |
| add action understanding alignment objective (ablation) | LinkVLA · Action Dreaming | mean success rate (%): 81.87 -> 87.16 | Improved |
| add action tokenization (ablation) | LinkVLA · Action Dreaming | mean success rate (%): 70.11 -> 81.63 | Improved |
| add C2F generation (ablation) | LinkVLA · Action Dreaming | mean success rate (%): 81.63 -> 81.87 | No clear effect |
| add alignment objective (ablation) | LinkVLA · DriveLM-VQA | SPICE: 71.3 -> 73.0 | Improved |
| add alignment objective (ablation) | LinkVLA · DriveLM commentary | SPICE: 53.6 -> 57.4 | Improved |
| use coarse-to-fine instead of auto-regressive decoding (ablation) | LinkVLA · Bench2Drive | latency (ms): 361ms -> 48ms | Improved |
| add spatial soft labeling (ablation) | LinkVLA · Bench2Drive | Success Rate (%): 72.73 -> 74.55 | Improved |
| add spatial soft labeling (ablation) | LinkVLA · Bench2Drive | Driving Score: 90.85 -> 91.01 | No clear effect |
| use navigation command instead of GPS target point (ablation) | LinkVLA · Bench2Drive | Driving Score: 91.01 -> 91.25 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- SimLingo — SimLingo: Vision-Only Closed-Loop Autonomous Driving with Language-Action Alignment (CVPR 2025) · better on 3 of 5 shared results
- Orion — ORION: A Holistic End-to-End Autonomous Driving Framework by Vision-Language Instructed Action Generation (ICCV 2025) · better on 3 of 3 shared results
- 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
- Less Language, More Latents: Annotation-Efficient VLAs for Driving (NeurIPS 2026 · similarity 0.94)
- AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (NeurIPS 2025 · similarity 0.93)
- Impromptu VLA: Open Weights and Open Data for Driving Vision-Language-Action Models (NeurIPS 2025 · similarity 0.91)
- LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving (NeurIPS 2026 · similarity 0.91)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- FineVLA: Fine-Grained Instruction Alignment for Steerable Vision-Language-Action Policies (NeurIPS 2026 · similarity 0.90)
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.
