---

📢 News

• [2026/5/20] We released code and checkpoints.
• [2026/5/20] We released our paper on arXiv.

📌 Table of Contents

• 🏛️ Model Zoo
• 🎯 Getting Started
• 📦 Data Preparation
  • Download Dataset
  • Set Up Configuration
  • Cache the Dataset
• ⚙️ Training and Evaluation
  • Evaluation
  • Training
• ❤️ Acknowledgements

🏛️ Model Zoo

Method	Backbone	Benchmark	PDMS	Weight Download
FlowPolicy	ResNet-34	NAVSIM	87.0	Google Drive
LTF*	ResNet-34	NAVSIM	88.7	Google Drive
LTFv7	ResNet-34	NAVSIM	89.8	Google Drive
MeanFuser + BeyondDrive	ResNet-34	NAVSIM	90.3	Google Drive
DiffusionDrive + BeyondDrive	ResNet-34	NAVSIM	89.2	Google Drive
WoTE + BeyondDrive	ResNet-34	NAVSIM	89.2	Google Drive

🎯 Getting Started

1. Clone BeyondDrive Repo

git clone https://github.com/wjl2244/BeyondDrive.git
cd BeyondDrive

2. Create Environment

conda create -n beyonddrive python=3.9 -y
conda activate beyonddrive
pip install -e .

📦 Data Preparation

NOTE: Please review and agree to the LICENSE file file before downloading the data.

1. Download Dataset

a. Download via NAVSIM offical installation.

Follow the instructions in the NAVSIM installation guide to download the dataset.

b. Download via Hugging Face

Alternatively, you can download the dataset using Hugging Face with the following commands:

export HF_ENDPOINT="https://huggingface.co"
# export HF_ENDPOINT="http://hf-mirror.com"  # Uncomment this line if you are in China

# Install the huggingface_hub tool
pip install -U "huggingface_hub"

# Download the OpenScene dataset
hf download --repo-type dataset OpenDriveLab/OpenScene --local-dir ./navsim_dataset/ --include "openscene-v1.1/*"

# Download the map data
cd download && ./download_maps.sh

2. Set Up Configuration

Move the download data to create the following structure.

navsim_workspace/
├── BeyondDrive/
├── dataset/
│    ├── maps/
│    ├── navsim_logs/
│    │   ├── test/
│    │   ├── trainval/
│    ├── sensor_blobs/
│    │   ├── test/
│    │   ├── trainval/
└── cache/
     ├── navtest_v1_metric_cache/
     └── traintest_v1_cache/

3. Cache the Dataset

We provide a script to cache the dataset and metrics.

cd BeyondDrive

# Cache the dataset. (navtrain and navtest)
bash scripts/evaluation/run_dataset_cache.sh

# Cache the metric.
bash scripts/evaluation/run_metric_cache.sh

⚙️ Training and Evaluation

1. Evaluation

(1) Please download the pre-trained checkpoints from here and place them in the navsim_workspace/BeyondDrive/exp/ directory.

(2) Please download the ResNet-34 pretrained weights from here. After downloading, update the corresponding path in the configuration file: navsim/agents/transfuser/transfuser_config.py, navsim/agents/flowpolicy/flowpolicy_config.py, navsim/agents/meanfuser/meanfuser_config.py, navsim/agents/diffusiondrive/diffusiondrive_config.py

cd BeyondDrive

# FlowPolicy Evaluation
bash scripts/evaluation/run_evaluation_flowpolicy.sh

# LTF* & LTFv7
bash scripts/evaluation/run_evaluation_transfuser.sh

# MeanFuser+BeyondDrive
bash scripts/evaluation/run_evaluation_meanfuser.sh

# DiffusionDrive+BeyondDrive
bash scripts/evaluation/run_evaluation_diffusiondrive.sh

Negative Samples Generate

Before training(LTFv7, DiffusionDrive+BeyondDrive and Meanfuser+BeyondDrive), negative dataset should be downloaded from Google Drive, or generated locally.

bash scripts/evaluation/run_generate_negative_samples_pool.sh

2. Training

cd BeyondDrive

# FlowPolicy Training
bash scripts/training/run_training_flowpolicy.sh

# LTF* & LTFv7
bash scripts/training/run_training_transfuser.sh

# MeanFuser+BeyondDrive
bash scripts/training/run_training_meanfuser.sh

# DiffusionDrive+BeyondDrive
bash scripts/training/run_training_diffusiondrive.sh

❤️ Acknowledgements

We acknowledge all the open-source contributors for the following projects to make this work possible:

• MeanFlow | NAVSIM | HUGSIM | DiffusionDrive | GTRS