Research map / Vision-language and multimodal / Robot action and navigation
Teaching Vision-Language-Action Models What to See and Where to Look
Yuguang Yang, Canyu Chen, Zhewen Tan, Yizhi Wang, Zichao Feng, Chunyang Liu, Kehua Sheng, Bo Zhang, Yan Wang, Juan Zhang, Linlin Yang, Baochang Zhang 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 DVD pretraining in place of VQA pretraining (ablation) | Qwen2.5-VL-3B planner with CoT · NAVSIM navtest | PDMS: 86.4 -> 87.1 | Improved |
| add 2D-TGP guidance (ablation) | DVD + CoT planner · NAVSIM navtest | PDMS: 87.1 -> 88.2 | Improved |
| add GRPO after TGP-guided CoT SFT (ablation) | DriveTeach-VLA · NAVSIM navtest | PDMS: 88.2 -> 90.4 | Improved |
| add 2D-TGP when GRPO is used (ablation) | DVD + CoT + GRPO planner · NAVSIM navtest | PDMS: 89.1 -> 90.4 | Improved |
| use block-wise (2x4) distillation instead of patch-by-patch (ablation) | DVD TGP-Prompter · NAVSIM navtest | PDMS: 86.5 -> 87.6 | Improved |
| use 4x7 blocks instead of 2x4 (ablation) | DVD TGP-Prompter · NAVSIM navtest | PDMS: 87.6 -> 87.1 | Hurt |
| add DVD pretraining to the TGP-Prompter (ablation) | TGP-Prompter · NAVSIM navtest | PDMS: 86.2 -> 87.6 | Improved |
| vary lambda_TGP from 0.05 to 0.10 (ablation) | DVD pretraining · NAVSIM navtest | PDMS: 86.3 -> 87.6 | Improved |
| raise lambda_TGP from 0.10 to 0.15 (ablation) | DVD pretraining · NAVSIM navtest | PDMS: 87.6 -> 87.4 | Hurt |
| drop 40% of bboxes in DVD distillation (ablation) | DVD TGP-Prompter · NAVSIM navtest | PDMS: 87.6 -> 85.3 | Hurt |
| jitter 20% of bbox corners in DVD (ablation) | DVD TGP-Prompter · NAVSIM navtest | PDMS: 87.6 -> 87.1 | Mixed |
| replace dual models with a single multi-turn QA model (ablation) | Prompter-Planner · NAVSIM navtest | PDMS: 88.2 -> 87.0 | Hurt |
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.
- AutoVLA — AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (NeurIPS 2025) · better on 3 of 3 shared results
- ReCogDrive — ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Driving (ICLR 2026) · 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 0.92)
- Impromptu VLA: Open Weights and Open Data for Driving Vision-Language-Action Models (NeurIPS 2025 · similarity 0.92)
- AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (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)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · similarity 0.90)
- SOLVE: Synergy of Language-Vision and End-to-End Networks for Autonomous Driving (CVPR 2025 · 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.
