Research map / Vision-language and multimodal / Robot action and navigation
Senna: Bridging Large Vision-Language Models and End-to-End Autonomous Driving
Bo Jiang, Shaoyu Chen, Bencheng Liao, Xingyu Zhang, Wei Hong Yin, Qian Zhang, Chang Huang, Wenyu Liu, Xinggang Wang
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 meta-action planning QA to training (ablation) | Senna · DriveX val | planning accuracy (%): 52.67 -> 71.21 | Improved |
| add scene description QA to training (ablation) | Senna · DriveX val | planning accuracy (%): 53.65 -> 71.21 | Improved |
| add traffic light QA to training (ablation) | Senna · DriveX val | planning accuracy (%): 57.39 -> 71.21 | Improved |
| add planning explanation QA to training (ablation) | Senna · DriveX val | planning accuracy (%): 67.91 -> 71.21 | Improved |
| add VRU identification QA to training (ablation) | Senna · DriveX val | planning accuracy (%): 68.44 -> 71.21 | Improved |
| add motion intention prediction QA to training (ablation) | Senna · DriveX val | planning accuracy (%): 68.70 -> 71.21 | Improved |
| use surround-view multi-image input instead of front view only (ablation) | Senna · DriveX val | planning accuracy (%): 64.91 -> 71.21 | Improved |
| compress each image to 128 tokens instead of the original 576 (ablation) | Senna · DriveX val | planning accuracy (%): 0.00 -> 71.21 | Improved |
| reduce image tokens from 128 to 64 or 32 per image (ablation) | Senna · DriveX val | planning accuracy (%): 71.21 -> 70.42 -> 69.36 | No clear effect |
| scale training data from 100k to 800k clips (ablation) | Senna · DriveX val | planning accuracy (%): 57.45 -> 71.21 | Improved |
| use mix pre-training with driving fine-tuning and planning fine-tuning (ablation) | Senna · DriveX val | planning accuracy (%): 60.01 -> 71.21 | Improved |
| predict meta-actions with LVLM instead of low-frequency trajectories as connector | Senna vs DriveVLM · DriveX val | L2 error at 3s (m): 3.00 -> 2.62 | Improved |
Later papers that compared with this method
1 later accepted paper reports Senna-2-3B as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Intend, Reflect, Refine: An Adaptive Multimodal Reflection Framework for Autonomous Driving (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 1 of 1 on most shared results.
- VILA — VILA: Learning Image Aesthetics From User Comments With Vision-Language Pretraining (CVPR 2023) · 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
- Distilling Multi-modal Large Language Models for Autonomous Driving (CVPR 2025 · similarity 0.91)
- SOLVE: Synergy of Language-Vision and End-to-End Networks for Autonomous Driving (CVPR 2025 · similarity 0.90)
- Less Language, More Latents: Annotation-Efficient VLAs for Driving (NeurIPS 2026 · similarity 0.90)
- AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (NeurIPS 2025 · similarity 0.90)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · similarity 0.90)
- VLP: Vision Language Planning for Autonomous Driving (CVPR 2024 · 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.
