LEAD: Minimizing Learner–Expert Asymmetry in End-to-End Driving

Project Page | Paper | Code

Official CARLA dataset accompanies our paper LEAD: Minimizing Learner–Expert Asymmetry in End-to-End Driving.

We release the complete pipeline required to achieve state-of-the-art closed-loop performance on the Bench2Drive benchmark. Built around the CARLA simulator, the stack features a data-centric design with:

• Extensive visualization suite and runtime type validation for easier debugging.
• Optimized storage format, packs 72 hours of driving in ~200GB.
• Native support for NAVSIM and Waymo Vision-based E2E and extending those benchmarks through closed-loop simulation and synthetic data for additional supervision during training.

Find more information on https://github.com/autonomousvision/lead.

Format

Each route is stored as a sequence of synchronized frames. All sensor modalities are ego-centric and time-aligned. In addition to the nominal sensor suite, we provide a second, perturbated sensor stack corresponding to a counterfactual ego state used for recovery supervision.

├── bboxes/                  # Per-frame 3D bounding boxes for all actors
├── depth/                   # Compressed depth maps (should be used for auxiliary supervision only)
├── depth_perturbated        # Depth from a perturbated ego state
├── hdmap/                   # Ego-centric rasterized HD map
├── hdmap_perturbated        # HD map aligned to perturbated ego pose
├── lidar/                   # LiDAR point clouds
├── metas/                   # Per-frame metadata and ego state
├── radar/                   # Radar detections
├── radar_perturbated        # Radar detections from perturbated ego state
├── rgb/                     # Front-facing RGB images
├── rgb_perturbated          # RGB images from perturbated ego state
├── semantics/               # Semantic segmentation maps
├── semantics_perturbated    # Semantics from perturbated ego state
└── results.json             # Route-level summary and evaluation metadata

Download

You can either download a single route (useful for quick inspection / debugging) or clone the full dataset via Git LFS and unzip all routes.

Note: Download the dataset after setting up the lead repository.

Option 1: Download a single route

bash scripts/download_one_route.sh

Option 2: Download all routes (Git LFS)

Clone the dataset repository directly into the expected directory:

git lfs install
git clone https://huggingface.co/datasets/ln2697/lead_carla data/carla_leaderboard2/zip

Unzip routes

Run

bash scripts/unzip_routes.sh

Citation

If you find this work useful, please cite:

@article{Nguyen2025ARXIV,
  title={LEAD: Minimizing Learner-Expert Asymmetry in End-to-End Driving},
  author={Nguyen, Long and Fauth, Micha and Jaeger, Bernhard and Dauner, Daniel and Igl, Maximilian and Geiger, Andreas and Chitta, Kashyap},
  journal={arXiv preprint arXiv:2512.20563},
  year={2025}
}

License

This project is released under the MIT License