Research map / Vision-language and multimodal / Multimodal LLMs
Reason2Drive: Towards Interpretable and Chain-based Reasoning for Autonomous Driving
Ming Nie, Renyuan Peng, Chunwei Wang, Xinyue Cai, Jianhua Han, Hang Xu, Li Zhang
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 reasoning-chain tasks to instruction tuning data (ablation) | Reason2Drive model · Reason2Drive val | ADRScore: 0.282 -> 0.351 | Improved |
| add perception task data on top of reasoning (ablation) | Reason2Drive model · Reason2Drive val | ADRScore: 0.351 -> 0.407 | Improved |
| add prediction task data on top of perception and reasoning (ablation) | Reason2Drive model · Reason2Drive val | ADRScore: 0.407 -> 0.463 | Improved |
| use multi-frame video features instead of single-frame image features (ablation) | Reason2Drive model · Reason2Drive val | ADRScore-S: 0.379 -> 0.394 | Improved |
| add region-level features from the prior tokenizer (ablation) | Reason2Drive model · Reason2Drive val | ADRScore-S: 0.394 -> 0.418 | Improved |
| add positional embedding to perceptual priors (ablation) | Reason2Drive model · Reason2Drive val | ADRScore-S: 0.418 -> 0.432 | Improved |
| pre-train the instructed vision decoder (ablation) | Reason2Drive model · Reason2Drive val | ADRScore-S: 0.361 -> 0.396 | Improved |
| add text embedding input to the vision decoder (ablation) | Reason2Drive model · Reason2Drive val | ADRScore-S: 0.396 -> 0.425 | Improved |
| use ViT-G/14 EVA-CLIP encoder instead of ViT-L/14 (ablation) | Reason2Drive model · Reason2Drive val | ADRScore-S: 0.397 -> 0.432 | Improved |
| pre-train on Reason2Drive before fine-tuning for control signals | InstructBLIP · nuScenes control signals | Speed RMSE: 3.743 -> 3.086 | Improved |
| replace textual perception priors and LLM-only decoding with prior tokenizer and instructed vision decoder | InstructBLIP Vicuna-7B · Reason2Drive val | ADRScore-S: 0.214 -> 0.432 | Improved |
| train on nuScenes only and test on unseen Waymo+ONCE | Ours · Reason2Drive generalization | ADRScore-S: 0.443 -> 0.385 | Mixed |
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.
- MiniGPT-4 — MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models (ICLR 2024) · 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
- nuReasoning: A Reasoning-Centric Dataset and Benchmark for Long-Tail Autonomous Driving (NeurIPS 2026 · similarity 0.92)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · similarity 0.91)
- DriveCombo: Benchmarking Compositional Traffic Rule Reasoning in Autonomous Driving (CVPR 2026 · similarity 0.91)
- DriveSpatial: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving (NeurIPS 2026 · similarity 0.90)
- STSBench: A Spatio-temporal Scenario Benchmark for Multi-modal Large Language Models in Autonomous Driving (NeurIPS 2025 · similarity 0.90)
- SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving (CVPR 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.
