fsdrive

Official code for FutureSightDrive, a NeurIPS 2025 spotlight research paper on autonomous driving. It has the model think visually about future scenes before deciding how to drive.

Browse open issues

Pick a small issue you can reproduce, then fix it with your AI coding agent.

Share on XNo license declared

Overview

FutureSightDrive, or FSDrive, is the official code for a research paper accepted as a NeurIPS 2025 spotlight. It proposes a spatio-temporal chain of thought that lets a vision-language-action model for end-to-end autonomous driving think visually when planning a trajectory. According to the authors, it unifies visual generation and understanding with minimal data. The repository covers installation, data preparation, training, inference, evaluation and visualization.

Key features

  • Spatio-temporal chain of thought for driving models
  • Visual reasoning for trajectory planning
  • Unifies visual generation and understanding
  • Guides for training, inference, evaluation and visualization

Best for

Researchers working on autonomous driving, vision-language-action models or world models who want to reproduce or build on this paper. It is research code, not a ready driving product.

Upstream
MIV-XJTU/FSDrive
Fork on GitHub
Guo-astro/fsdrive
Upstream stars
832
Category
Finance, data and research
Language
Python
License
No license declaredWithout a license, the author keeps all rights. Ask the upstream owner before reusing the code.
Forked
2025-11-09
Sync status
In syncLast synced 2026-10-19