Bench2Drive

Bench2Drive is an autonomous driving benchmark based on the CARLA leaderboard 2.0. It consists of 220 short routes featuring safety critical scenarios. The evaluation is performed closed-loop in the CARLA simulator. The performance of an entire driving stack is being evaluated.

Model Name Driving Score ↑ Paper Title Repository
AlignDrive 89.07 📄 AlignDrive: Aligned Lateral-Longitudinal Planning for End-to-End Autonomous Driving -
BridgeDrive 86.87 📄 BridgeDrive: Diffusion Bridge Policy for Closed-Loop Trajectory Planning in Autonomous Driving -
HiP-AD 86.77 📄 HiP-AD: Hierarchical and Multi-Granularity Planning with Deformable Attention for Autonomous Driving in a Single Decoder
R2SE 86.28 📄 Reinforced Refinement with Self-Aware Expansion for End-to-End Autonomous Driving -
SimLingo-Base (CarLLaVa) 85.94 📄 CarLLaVA: Vision language models for camera-only closed-loop driving
TransFuser++ 84.21 📄 Hidden Biases of End-to-End Driving Models
GaussianFusion 79.40 📄 GaussianFusion: Gaussian-Based Multi-Sensor Fusion for End-to-End Autonomous Driving
PGS 78.08 📄 Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving -
ORION 77.70 📄 ORION: A Holistic End-to-End Autonomous Driving Framework by Vision-Language Instructed Action Generation
VLR-Driver 75.01 📄 VLR-Driver: Large Vision-Language-Reasoning Models for Embodied Autonomous Driving -
Raw2Drive 74.36 📄 Raw2Drive: Reinforcement Learning with Aligned World Models for End-to-End Autonomous Driving (in CARLA v2) -
ETA 74.33 📄 ETA: Efficiency through Thinking Ahead, A Dual Approach to Self-Driving with Large Models
DriveMoE 74.22 📄 DriveMoE: Mixture-of-Experts for Vision-Language-Action Model in End-to-End Autonomous Driving -
Hydra-NeXt 73.86 📄 Hydra-NeXt: Robust Closed-Loop Driving with Open-Loop Training
VL (on failure) 73.29 📄 Validity Learning on Failures: Mitigating the Distribution Shift in Autonomous Vehicle Planning -
ReCogDrive 71.36 📄 ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Driving
DRIVER 68.90 -
DiffAD 67.92 📄 DiffAD: A Unified Diffusion Modeling Approach for Autonomous Driving -
NavigationDrive 67.17 -
VDRive 66.25 📄 VDRive: Leveraging Reinforced VLA and Diffusion Policy for End-to-end Autonomous Driving -
iPad 65.02 📄 iPad: Iterative Proposal-centric End-to-End Autonomous Driving
DriveAdapter 64.22 📄 DriveAdapter: Breaking the Coupling Barrier of Perception and Planning in End-to-End Autonomous Driving
ReasonPlan 64.01 📄 ReasonPlan: Unified Scene Prediction and Decision Reasoning for Closed-loop Autonomous Driving
Drivetransformer-Large 63.46 📄 DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Driving
ThinkTwice 62.44 📄 Think Twice before Driving: Towards Scalable Decoders for End-to-End Autonomous Driving
TCP-traj 59.90 📄 Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baseline
SpaRC-AD 55.6 📄 SpaRC-AD: A Baseline for Radar-Camera Fusion in End-to-End Autonomous Driving -
DiFSD 52.02 📄 DiFSD: Ego-Centric Fully Sparse Paradigm with Uncertainty Denoising and Iterative Refinement for Efficient End-to-End Self-Driving
X-Driver 51.70 📄 X-Driver: Explainable Autonomous Driving with Vision-Language Models -
TCP-traj w/o distillation 49.30 📄 Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baseline
CogAD 48.30 📄 CogAD: Cognitive-Hierarchy Guided End-to-End Autonomous Driving -
MomAD 47.91 📄 Don't Shake the Wheel: Momentum-Aware Planning in End-to-End Autonomous Driving
UniAD-Base 45.81 📄 Planning-oriented Autonomous Driving
TTOG 45.23 📄 Two Tasks, One Goal: Uniting Motion and Planning for Excellent End To End Autonomous Driving Performance -
GenAD 44.81 📄 GenAD: Generative End-to-End Autonomous Driving
SparseDrive 44.54 📄 SparseDrive: End-to-End Autonomous Driving via Sparse Scene Representation
VAD 42.35 📄 VAD: Vectorized Scene Representation for Efficient Autonomous Driving
UniAD-Tiny 40.73 📄 Planning-oriented Autonomous Driving
TCP 40.70 📄 Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baseline
VAD + SERA 35.64 📄 From Failures to Fixes: LLM-Driven Scenario Repair for Self-Evolving Autonomous Driving -
TCP-ctrl 30.47 📄 Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baseline
AD-MLP 18.05 📄 Bench2Drive: Towards Multi-Ability Benchmarking of Closed-Loop End-To-End Autonomous Driving