MindDriver: Introducing Progressive Multimodal Reasoning for Autonomous Driving

🎉🎉CVPR 2026 🎉🎉

Lingjun Zhang^1*, Yujian Yuan^1,2*, Changjie Wu^1†, Xinyuan Chang¹, Xin Cai³, Shuang Zeng^1,4, Linzhe Shi¹, Sijin Wang¹, Hang Zhang¹, Mu Xu¹,

¹Amap, Alibaba Group, ²The Hong Kong University of Science and Technology, ³The Chinese University of Hong Kong, ⁴Xi'an Jiaotong University

(*) Equal contribution. (†) Project leader.

Comparison of different reasoning methods. Text reasoning struggles with space misalignment, while image reasoning suffers from guideless image prediction. Our proposed progressive multimodal reasoning conducts aligned smooth reasoning.

MindDriver: The proposed multimodal reasoning framework that enables VLM to imitate human-like progressive thinking for autonomous driving. MindDriver presents semantic understanding, semantic-to-physical space imagination, and physical-space trajectory planning.

🗓️ TODO

• Release MindDriver reasoning code
• Release whole MindDriver code
• Release checkpoints

🙏 Acknowledgement

Our work is primarily based on the following codebases:FSDrive, LLaMA-Factory, MoVQGAN, GPT-Driver, Agent-Driver. We are sincerely grateful for their work.