Awesome Autonomous Driving Research

A curated research collection for end-to-end autonomous driving, driving world models, vision-language-action models, recursive self-improvement, public datasets, and reproducible evaluation. The catalog currently contains 244 papers across four research tracks.

Last updated: 2026-10-11. Verify paper metadata against the latest arXiv or publisher version before citation.

Table of Contents

• Papers
  • End-to-End Autonomous Driving
  • Vision-Language-Action Models
  • Driving World Models
  • Recursive Self-Improvement for Autonomous Driving
• Public Datasets
  • Perception and Multimodal Understanding
  • Motion, Planning, and Cooperative Driving
  • Simulation and Synthetic Data
• Public Evaluation Benchmarks
  • Closed-Loop Planning
  • Open-Loop and Non-Reactive Planning
  • VLA and Driving Reasoning
  • Reliability and Robustness
• Leaderboard
• License

Papers

Tags were checked against paper titles, abstracts, and method descriptions. GitHub repositories were checked against paper titles, arXiv identifiers, official project pages, and repository contents. Code pending marks an announced implementation that is not yet available; Partial code identifies a limited release such as inference or data tools. Project-only repositories are identified in the resource notes. Missing links mean no matching implementation was verified as of the audit date. The per-paper evidence and search coverage are recorded in data/github-audit.json.

The RSI track groups autonomous-driving research with an explicit iterative improvement loop. Papers may describe this as self-improving, self-evolving, safe online learning, or agentic evolution; inclusion does not imply that a fully general recursively self-modifying system has been demonstrated.

End-to-End Autonomous Driving

77 papers; 52 with a verified public code link.

2026

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
NavSafe-∞ NavSafe-∞: Benchmarking Closed-Loop Driving Safety in Photorealistic Environments	2026	Closed-Loop Benchmark · Photorealistic Simulation · Safety Evaluation	💻
Driving on Registers, Reasoning on Risk Driving on Registers, Reasoning on Risk: Risk-Aware Occupancy for Register-Based End-to-End Autonomous Driving	2026	Occupancy Representation · Token Representation · Risk-Aware Planning
DRiF Data-Driven Risk Fields for Safer End-to-End Autonomous Driving	2026	Relative Risk Supervision · BEV Risk Fields · Safety-Aware Planning
RUDC A Risk-Sensitive and Uncertainty-Aware Decision-Making and Control Framework for Safe and Robust Autonomous Driving	2026	Safe Reinforcement Learning · Policy Uncertainty · Control Barrier Functions
AD-Diff Bench Understanding Autonomous Driving Datasets by Describing Differences between Image Subsets in Natural Language	2026	Dataset Understanding · Benchmark · Domain Shift	💻
CrashDiffuser CrashDiffuser: VLM-Guided Collision Intent Reasoning for Fine-Grained Safety-Critical Traffic Scenario Generation	2026	Scenario Generation · VLM-Guided Reasoning · Safety Evaluation
NAVSIM Score Basis Audit When Shared Rollouts Fail in Defensive Driving Evaluation: A NAVSIM Score Basis Audit	ECCV 2026 Workshop	Evaluation Audit · Numerical Robustness · Defensive Driving	💻
NeoAD Drive Like Humans, Plan Like Machines: An Explicit Sense and Safety Aware Autonomous Driving Framework	CVPR 2026 Workshop	Explicit Planning Tokens · Safety-Aware Planning · BEV Representation	💻
DiDrive DiDrive: A Risk-Aware Hierarchical Diffusion Framework for Safe Offline Reinforcement Learning in Autonomous Driving	2026	Offline Reinforcement Learning · Risk-Aware Diffusion · Candidate Ranking
PriorEye PriorEye: Geospatial Visual Priors for End-to-End Autonomous Driving	ECCV 2026	Geospatial Priors · Anticipatory Planning	💻 🌐
FlowR2A FlowR2A: Learning Reward-to-Action Distribution for Multimodal Driving Planning	2026	Reward-Conditioned Planning · Flow Matching · Multi-Modal Trajectories	💻 🌐
World Engine World Engine: Towards the Era of Post-Training for Autonomous Driving	2026	Post-Training · Interactive Simulation · Long-Tail Safety	💻 🌐
Gigapixel Scaling Self-Play for End-to-End Driving	2026	Self-Play · Closed-Loop Reinforcement Learning	💻 Code pending: The repository contains a README and assets. The announced code, data, and model release schedule starts after this audit date.
COMPACT-VA Planning-aligned Token Compression for Long-Context Autonomous Driving	2026	Token Compression · Long-Context Memory · Planning-Aware Learning
TOAD Test-Time Trajectory Optimization for Autonomous Driving	2026	Test-Time Optimization · Cross-Entropy Search · Trajectory Scoring	💻 🌐
NTR NTR: Neural Token Reconstruction for Scene Token Bottleneck in End-to-End Driving	2026	Scene Tokenization · Perception-Free Planning · Reconstruction Supervision
CLOVER CLOVER: Closed-Loop Value Estimation and Ranking for End-to-End Autonomous Driving Planning	2026	Closed-Loop Self-Distillation · Trajectory Scoring · Multi-Metric Planning	💻 Partial code: Inference scripts and checkpoints are public; official training and pseudo-expert generation code remain pending.
123D 123D: Unifying Multi-Modal Autonomous Driving Data at Scale	2026	Driving Data Library · Multi-Modal Datasets · Data Standardization	💻
ReflectDrive-2 ReflectDrive-2: Reinforcement-Learning-Aligned Self-Editing for Discrete Diffusion Driving	2026	Discrete Diffusion · Trajectory Editing
RAD-2 RAD-2: Scaling Reinforcement Learning in a Generator-Discriminator Framework	2026	Diffusion Planning · Reinforcement Learning · Closed-Loop Adaptation	💻 🌐 Official shared RAD/RAD-2 repository. Released scripts are labeled RAD core components; a RAD-2-specific implementation was not verified.
MOSAIC Scaling-Aware Data Selection for End-to-End Autonomous Driving Systems	CVPR 2026	Data Selection · Data Scaling · Evaluation Alignment
FlowAD FlowAD: Ego-Scene Interactive Modeling for Autonomous Driving	2026	Ego-Scene Interaction · Scene Flow · Closed-Loop Evaluation	💻
SparseDriveV2 SparseDriveV2: Scoring is All You Need for End-to-End Autonomous Driving	2026	Trajectory Scoring · Sparse Proposals · Multi-Modal Planning	💻
HDP Unleashing the Potential of Diffusion Models for End-to-End Autonomous Driving	2026	Diffusion Planning · Reinforcement Learning · Real-World Validation	💻 🌐
MeanFuser MeanFuser: Fast One-Step Multi-Modal Trajectory Generation and Adaptive Reconstruction via MeanFlow for End-to-End Autonomous Driving	CVPR 2026	Flow Matching · One-Step Planning · Multi-Modal Trajectories	💻
PCM A Generalizable Physics-guided Causal Model for Trajectory Prediction in Autonomous Driving	ICRA 2026	Causal Modeling · Motion Prediction · Domain Generalization	💻
PlannerRFT PlannerRFT: Reinforcing Diffusion Planners through Closed-Loop and Sample-Efficient Fine-Tuning	2026	Reinforcement Fine-Tuning · Diffusion Planning · Adaptive Trajectories
DrivoR Driving on Registers	2026	Register Tokens · Vision Transformer Backbone · Trajectory Scoring	💻 🌐
AlignDrive AlignDrive: Aligned Lateral-Longitudinal Planning for End-to-End Autonomous Driving	2026	Lateral-Longitudinal Planning · Path-Conditioned Planning	💻 🌐 Code pending: The linked author repository currently contains only .gitattributes; implementation is not released.

2025

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
LEAD LEAD: Minimizing Learner-Expert Asymmetry in End-to-End Driving	2025	Imitation Learning · Privileged Learning · Closed-Loop Evaluation	💻
DiffusionDriveV2 DiffusionDriveV2: Reinforcement Learning-Constrained Truncated Diffusion Modeling in End-to-End Autonomous Driving	2025	Diffusion Planning · Reinforcement Learning · Diverse Trajectories	💻
Spatial Retrieval AD Spatial Retrieval Augmented Autonomous Driving	2025	Spatial Retrieval · Geospatial Priors · Robust Planning	💻 🌐
LAP LAP: Fast LAtent Diffusion Planner for Autonomous Driving	2025	Latent-Space Diffusion · Intent-Kinematics Disentanglement · Feature Alignment	💻 Code pending: The repository contains only a README and announces a future code release.
GuideFlow GuideFlow: Constraint-Guided Flow Matching for Planning in End-to-End Autonomous Driving	2025	Flow Matching · Safety-Constrained Planning	💻 Code pending: The repository contains documentation and images; the README says the code is still being organized.
DiffRefiner DiffRefiner: Coarse to Fine Trajectory Planning via Diffusion Refinement with Semantic Interaction for End to End Autonomous Driving	AAAI 2026	Diffusion Planning · Trajectory Refinement
MiMo-Embodied MiMo-Embodied: X-Embodied Foundation Model Technical Report	2025	Embodied Foundation Model · Cross-Embodiment Transfer · Autonomous Driving	💻 🌐
UniMM-V2X UniMM-V2X: MoE-Enhanced Multi-Level Fusion for End-to-End Cooperative Autonomous Driving	2025	Cooperative Driving · Multi-Agent Planning · V2X	💻
UniLION UniLION: Towards Unified Autonomous Driving Model with Linear Group RNNs	2025	LiDAR · Multi-View Fusion · Efficient Transformers	💻
ZTRS Zero-Human Demonstration End-to-end Autonomous Driving with Trajectory Scorer	2025	Trajectory Scoring · Zero-Shot Planning · Policy Optimization	💻
ResAD ResAD: Normalized Residual Trajectory Modeling for End-to-End Autonomous Driving	CVPR 2026	Residual Trajectory Modeling · Inertial Reference · Diffusion Planning	💻 🌐 Code pending: The author-linked repository currently contains only a README; implementation is not released.
EvaDrive EvaDrive: Evolutionary Adversarial Policy Optimization for End-to-End Autonomous Driving	2025	Evolutionary Reinforcement Learning · Iterative Planning · Closed-Loop Driving
DistillDrive DistillDrive: End-to-End Multi-Mode Autonomous Driving Distillation by Isomorphic Hetero-Source Planning Model	2025	Knowledge Distillation · Planning-Centric Learning	💻
DiffSemanticFusion DiffSemanticFusion: Semantic Raster BEV Fusion for Autonomous Driving via Online HD Map Diffusion	2025	HD Map Generation · BEV Semantics · Sensor Fusion	💻 Partial code: Mapless QCNet training and evaluation components are public; the full DiffSemanticFusion release is still being organized.
GEMINUS GEMINUS: Dual-aware Global and Scene-Adaptive Mixture-of-Experts for End-to-End Autonomous Driving	2025	Mixture-of-Experts · Driving Skills · Adaptive Planning	💻 Partial code: Data preprocessing tools are public; the training implementation remains pending.
TrajFlow TrajFlow: Multi-modal Motion Prediction via Flow Matching	IROS 2025	Flow Matching · Motion Prediction · Multi-Modal Forecasting	💻 🌐
ETA ETA: Efficiency through Thinking Ahead, A Dual Approach to Self-Driving with Large Models	ICCV 2025	Dual-System VLM · Fast-Slow Planning · Inference Efficiency	💻
GTRS Generalized Trajectory Scoring for End-to-end Multimodal Planning	CVPR 2025	Trajectory Scoring · Multi-Modal Planning	💻
DriveSuprim DriveSuprim: Towards Precise Trajectory Selection for End-to-End Planning	AAAI 2026	Trajectory Selection · Safety Scoring · Imitation Learning	💻
NAVSIM Pseudo-Simulation for Autonomous Driving	CORL 2025	Evaluation Benchmark · Pseudo-Simulation · Open-Loop Metrics	💻
GaussianFusion GaussianFusion: Gaussian-Based Multi-Sensor Fusion for End-to-End Autonomous Driving	NEURIPS 2025	Multi-Sensor Fusion · Gaussian Representation · End-to-End Planning	💻
iPad iPad: Iterative Proposal-centric End-to-End Autonomous Driving	2025	Proposal-Centric Planning · Iterative Refinement · Sparse Representation	💻
TransDiffuser TransDiffuser: Diverse Trajectory Generation with Decorrelated Multi-modal Representation for End-to-end Autonomous Driving	2025	Diffusion Planning · Multi-Modal Trajectories · Trajectory Diversity
BridgeAD Bridging Past and Future: End-to-End Autonomous Driving with Historical Prediction and Planning	CVPR 2025	Temporal Memory · Historical Prediction · Trajectory Planning	💻
Hydra-MDP++ Hydra-MDP++: Advancing End-to-End Driving via Expert-Guided Hydra-Distillation	2025	Knowledge Distillation · Expert-Guided Planning · Multi-Head Policy
TrajHF Learning Personalized Driving Styles via Reinforcement Learning from Human Feedback	2025	Human Feedback · Personalized Driving Styles · Trajectory Generation
HiP-AD HiP-AD: Hierarchical and Multi-Granularity Planning with Deformable Attention for Autonomous Driving in a Single Decoder	2025	Hierarchical Planning · Multi-Granularity Queries	💻
DriveTransformer DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Driving	ICLR 2025	Unified Perception-Prediction-Planning · Transformer Architecture	💻
GoalFlow GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation in End-to-End Autonomous Driving	2025	Flow Matching · Goal-Conditioned Planning · Multi-Modal Trajectories	💻
MomAD Don't Shake the Wheel: Momentum-Aware Planning in End-to-End Autonomous Driving	2025	Trajectory Momentum · Occlusion Robustness · End-to-End Planning	💻

2024

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
DiffusionDrive DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving	CVPR 2025	Diffusion Policy · Real-Time Planning · Multi-Modal Trajectories	💻
NAVSIM NAVSIM: Data-Driven Non-Reactive Autonomous Vehicle Simulation and Benchmarking	NEURIPS 2024	Benchmarking · Non-Reactive Simulation · Data-Driven Evaluation	💻
Hydra-MDP Hydra-MDP: End-to-end Multimodal Planning with Multi-target Hydra-Distillation	CVPR 2024	Knowledge Distillation · Multi-Teacher Learning · Trajectory Selection	💻 Code pending: The official repository contains a README and assets; code and model release are delayed.
SparseDrive SparseDrive: End-to-End Autonomous Driving via Sparse Scene Representation	2024	Sparse Scene Representation · Multi-Task Learning · End-to-End Planning	💻
VADv2 VADv2: End-to-End Vectorized Autonomous Driving via Probabilistic Planning	2024	Vectorized Scene Representation · End-to-End Planning	💻 🌐 Partial code: Core VADv2 model and configuration are public in the VAD repository and require integration with VADv1.

2023

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
BEV-Planner Is Ego Status All You Need for Open-Loop End-to-End Autonomous Driving?	CVPR 2024	BEV Representation · Trajectory Planning	💻
VAD VAD: Vectorized Scene Representation for Efficient Autonomous Driving	ICCV 2023	Vectorized Scene Representation · End-to-End Planning	💻

2022

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
UniAD Planning-oriented Autonomous Driving	CVPR 2023	Unified Autonomous Driving · Planning-Oriented Learning	💻 🌐
ST-P3 ST-P3: End-to-end Vision-based Autonomous Driving via Spatial-Temporal Feature Learning	ECCV 2022	Spatial-Temporal Learning · End-to-End Planning · Closed-Loop Evaluation	💻 Official repository transferred from OpenPerceptionX to OpenDriveLab.
TransFuser TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Driving	2022	Sensor Fusion · Imitation Learning · End-to-End Driving	💻

2021

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
NEAT NEAT: Neural Attention Fields for End-to-End Autonomous Driving	ICCV 2021	Neural Attention Fields · Bird’s-Eye View · Imitation Learning	💻
Roach End-to-End Urban Driving by Imitating a Reinforcement Learning Coach	ICCV 2021	Imitation Learning · Reinforcement Learning · CARLA	💻
IVMP Learning Interpretable End-to-End Vision-Based Motion Planning for Autonomous Driving with Optical Flow Distillation	ICRA 2021	Vision-Based Planning · Optical Flow Distillation · Interpretability	🌐
TransFuser Multi-Modal Fusion Transformer for End-to-End Autonomous Driving	CVPR 2021	Transformer · Sensor Fusion · Imitation Learning	💻
MP3 MP3: A Unified Model to Map, Perceive, Predict and Plan	CVPR 2021	Mapless Driving · Online Mapping · Neural Planning

2020

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
P3 Perceive, Predict, and Plan: Safe Motion Planning Through Interpretable Semantic Representations	ECCV 2020	Semantic Occupancy · Joint Planning · Interpretability
Data Aggregation Exploring Data Aggregation in Policy Learning for Vision-Based Urban Autonomous Driving	CVPR 2020	DAgger · Imitation Learning · Policy Generalization	💻
Implicit Affordances End-to-End Model-Free Reinforcement Learning for Urban Driving Using Implicit Affordances	CVPR 2020	Reinforcement Learning · Implicit Affordances · Urban Driving	💻 🌐 Partial code: Official inference code and model weights; the repository is a fork of Learning by Cheating.

Vision-Language-Action Models

75 papers; 44 with a verified public code link.

2026

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
FocusDrive FocusDrive: Reasoning with Visual Focus for Autonomous Driving	2026	Visual Focus · Grounded Reasoning · Trajectory Planning
iDriveVLA Evaluation Is All You Need for Multi-Modal Autonomous Driving	2026	Multi-Modal Planning · Trajectory Scoring · VLM Guidance
Run-then-Walk Sometimes You Gotta Run Before You Can Walk: Run-then-Walk Scheduling Strategy for VLM Autonomous Driving	2026	VLM Planning · Reinforcement Learning · Safety
FIVE-VLA FIVE-VLA: Fast and EffectIVE Autonomous Driving with Recurrent Action Memory	2026	Efficient VLA · Recurrent Action Memory · Direct Trajectory Prediction
Qwen-Drive-1.0 Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving	2026	Vision-Language Foundation Model · 3D Perception · Joint Planning	💻
HybridDriveVLA HybridDriveVLA: Vision-Language-Action Model with Visual CoT reasoning and ToT Evaluation for Autonomous Driving	2026	Visual Chain-of-Thought · Waypoint Evaluation · Safety and Comfort
Plug-and-Play Plug-and-Play Traffic Element Awareness for End-to-End Autonomous Driving	2026	Traffic Element Conditioning · End-to-End Planning	💻
BrainWAM BrainWAM: Action-Space Coordination of Semantic Priors and Predictive Dynamics for Autonomous Driving	2026	VLA-WAM Integration · Predictive Dynamics · Joint Planning
DriveVLA-M0 DriveVLA-M0: Failure-Aware Memory Augmentation for Autonomous Driving	2026	Failure Memory · Retrieval Augmentation · Distribution Shift	💻
FactorDrive FactorDrive: Adaptive Multi-Step Reasoning Driven by Planning-Critical Factors for End-to-End Autonomous Driving	2026	Spatial Reasoning · Reasoning Optimization · Trajectory Planning
DEFT-RLVR Deferred Exposure of Future Trajectories for Verifiable Reasoning in Autonomous Driving VLMs	2026	Reinforcement Learning with Verifiable Rewards · Reasoning Supervision	💻
WorkDrive WorkDrive: Roadwork Chain of Causation for Autonomous Driving	2026	Work-Zone Reasoning · VLM Planning · Robustness
Post-Training Survey Post-Training in End-to-End Autonomous Driving	2026	Survey · Post-Training · Policy Optimization	💻 🌐 Official companion paper collection for the survey; no driving implementation is claimed.
LingBot-VLA 2.0 From Foundation to Application: Improving VLA Models in Practice	2026	Generalist VLA · Data Scaling · Real-World Deployment	💻 🌐
AnchorVLA AnchorVLA: Bridging Discrete Decisions and Continuous Trajectories for Vision-Language-Action Planning	2026	Navigation-Conditioned Planning · Continuous Actions
DriveTeach-VLA Teaching Vision-Language-Action Models What to See and Where to Look	ECCV 2026	Spatial Grounding · Action-Grounded Supervision	💻
WVM World Value Models for Robotic Manipulation	2026	Cross-Domain Robotics · World-Value Modeling
Qwen-RobotNav Qwen-RobotNav Technical Report: A Scalable Navigation Model Designed for an Agentic Navigation System	2026	Agentic Navigation · Configurable Visual Context · Cross-Domain Navigation	💻 🌐 Official report and demos are public; implementation and model weights are not released.
DriveReward DriveReward: A Comprehensive Dataset and Generative Vision-Language Reward Model for Autonomous Driving	2026	Reward Modeling · Trajectory Selection · Data Scaling
ChainFlow-VLA ChainFlow-VLA: Causal Flow Planning with Vision-Language Models	2026	Causal Flow Matching · Trajectory Generation · Global Consistency	💻
IKN Grounding Driving VLA via Inverse Kinematics	2026	Inverse Kinematics · Future-State Grounding · Trajectory Recovery
CRAFT CRAFT: Counterfactual-to-Interactive Reinforcement Fine-Tuning for Driving Policies	2026	Counterfactual Learning · Interactive Reinforcement Learning · Policy Fine-Tuning	💻 🌐
OneDrive OneDrive: Unified Multi-Paradigm Driving with Vision-Language-Action Models	2026	Unified Multimodal Decoding · Structured Actions · End-to-End Planning	💻 Partial code: nuScenes code and training scripts are public; model weights and the NAVSIM release remain pending.
UniDriveVLA UniDriveVLA: Unifying Understanding, Perception, and Action Planning for Autonomous Driving	2026	Joint Perception and Planning · 3D Grounding · VLA	💻
DVGT-2 DVGT-2: Vision-Geometry-Action Model for Autonomous Driving at Scale	2026	Vision-Geometry-Action · 3D Geometry · Action Planning	💻
AutoDrive-P3 AutoDrive-P3: Unified Chain of Perception-Prediction-Planning Thought via Reinforcement Fine-Tuning	ICLR 2026	Chain-of-Thought · Reinforcement Fine-Tuning · Unified Driving	💻 The original repository redirects here. This is a follow-up implementation that also incorporates Run-then-Walk.
Learning from Mistakes Learning from Mistakes: Post-Training for Driving VLA with Takeover Data	2026	Takeover Learning · Post-Training · Distribution Shift
AutoMoT AutoMoT: A Unified Vision-Language-Action Model with Asynchronous Mixture-of-Transformers for End-to-End Autonomous Driving	2026	Asynchronous VLA · Mixture-of-Transformers · Efficient Inference	💻 🌐
DynVLA DynVLA: Learning World Dynamics for Action Reasoning in Autonomous Driving	2026	World Dynamics Reasoning · Dynamics Tokens · Action Planning	💻 🌐
EvoDriveVLA EvoDriveVLA: Evolving Driving VLA Models via Collaborative Perception-Planning Distillation	2026	Knowledge Distillation · Perception-Planning Alignment	💻
Curious-VLA Devil is in Narrow Policy: Unleashing Exploration in Driving VLA Models	CVPR 2026	Exploration · Reinforcement Learning · Policy Improvement	💻
LaST-VLA LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving	2026	Latent Reasoning · Spatio-Temporal Grounding	💻 Code pending: Inference code, training code, and the training dataset remain unchecked on the release roadmap.
LinkVLA Unifying Language-Action Understanding and Generation for Autonomous Driving	2026	Language-Action Alignment · Parallel Action Generation
ELF-VLA Unleashing VLA Potentials in Autonomous Driving via Explicit Learning from Failures	2026	Learning from Failures · Reinforcement Learning · Long-Tail Exploration	💻 Code pending: Inference code, training code, and the training dataset remain unchecked on the release roadmap.
NoRD NoRD: A Data-Efficient Vision-Language-Action Model that Drives without Reasoning	CVPR 2026	Data Efficiency · Reasoning-Free VLA · Driving Policy	💻
VGGDrive VGGDrive: Empowering Vision-Language Models with Cross-View Geometric Grounding for Autonomous Driving	CVPR 2026	Cross-View Geometry · 3D Grounding · VLM	💻
HiST-VLA HiST-VLA: A Hierarchical Spatio-Temporal Vision-Language-Action Model for End-to-End Autonomous Driving	2026	Hierarchical Planning · Spatio-Temporal Reasoning · 3D Grounding
Dual-System Driving From Representational Complementarity to Dual Systems: Synergizing VLM and Vision-Only Backbones for End-to-End Driving	2026	Dual-System VLM · Vision-Only Backbones · Representation Complementarity
SparseOccVLA SparseOccVLA: Bridging Occupancy and Vision-Language Models via Sparse Queries for Unified 4D Scene Understanding and Planning	2026	Sparse Occupancy · 4D Scene Understanding · Unified Planning	💻
SGDrive SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving	2026	Scene-to-Goal Reasoning · Hierarchical Planning · 3D Grounding	💻
FLARE FLARE: Learning Future-Aware Latent Representations from Vision-Language Models for Autonomous Driving	2026	Latent Future Representation · VLM Transfer · Trajectory Planning

2025

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
ColaVLA ColaVLA: Leveraging Cognitive Latent Reasoning for Hierarchical Parallel Trajectory Planning in Autonomous Driving	CVPR 2026	Latent Reasoning · Hierarchical Planning · Parallel Trajectories	💻 🌐
SpaceDrive SpaceDrive: Infusing Spatial Awareness into VLM-based Autonomous Driving	2025	Spatial Awareness · 3D Reasoning · VLM Planning	💻
WAM-Flow WAM-Flow: Parallel Coarse-to-Fine Motion Planning via Discrete Flow Matching for Autonomous Driving	2025	Discrete Flow Matching · Parallel Planning · Structured Actions	💻 Partial code: SFT, inference, and evaluation code are public; reinforcement-learning code remains pending.
OpenREAD OpenREAD: Reinforced Open-Ended Reasoning for End-to-End Autonomous Driving with LLM-as-Critic	2025	Reasoning Fine-Tuning · Reinforcement Learning · Driving Policy	💻 Partial code: Models, inference, and evaluation scripts are public; training code remains pending.
Alpamayo-R1 Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail	2025	Causal Reasoning · Trajectory Planning · Long-Tail Driving	💻 🌐 Partial code: Core model and inference are public. This checkpoint uses supervised learning and does not include the paper's RL post-training weights.
Cosmos-Predict2.5 World Simulation with Video Foundation Models for Physical AI	2025	World Foundation Model · Video Generation · Flow-Based Modeling	💻
ReflectDrive-1 Discrete Diffusion for Reflective Vision-Language-Action Models in Autonomous Driving	2025	Discrete Diffusion · Reflective Reasoning · Trajectory Generation	💻 Code pending: The repository contains only a README and license; NAVSIM and training code are still pending.
OccVLA OccVLA: Vision-Language-Action Model with Implicit 3D Occupancy Supervision	2025	Occupancy Grounding · 3D Spatial Understanding · VLA
FastDriveVLA FastDriveVLA: Efficient End-to-End Driving via Plug-and-Play Reconstruction-based Token Pruning	AAAI 2026	Visual Token Pruning · Efficient Inference · VLA
AutoVLA AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning	NEURIPS 2025	Joint Reasoning and Action · Autoregressive VLA · Reinforcement Learning	💻 🌐
ReCogDrive ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Driving	2025	Cognitive Planning · Action Tokenization · Reinforcement Learning	💻
DiffVLA DiffVLA: Vision-Language Guided Diffusion Planning for Autonomous Driving	2025	Diffusion Planning · BEV Efficiency · Action Diversity	💻
DriveMoE DriveMoE: Mixture-of-Experts for Vision-Language-Action Model in End-to-End Autonomous Driving	2025	Mixture-of-Experts · Scenario Specialization · End-to-End Planning	💻 🌐
LightEMMA LightEMMA: A Longitudinal Evaluation of Vision-Language Models for Autonomous Driving	2025	VLM Evaluation · Open-Loop Benchmarking · Inference Cost	💻
OpenDriveVLA OpenDriveVLA: Towards End-to-end Autonomous Driving with Large Vision Language Action Model	2025	Spatially Grounded Actions · Language-Conditioned Driving · VLA	💻 Partial code: Model and inference code are public; training scripts remain pending.
Cosmos-Reason1 Cosmos-Reason1: From Physical Common Sense To Embodied Reasoning	2025	Physical Reasoning · Embodied Foundation Model	💻
SimLingo SimLingo: Vision-Only Closed-Loop Autonomous Driving with Language-Action Alignment	CVPR 2025	Language-Action Alignment · Closed-Loop Driving · Action Dreaming	💻
AlphaDrive AlphaDrive: Unleashing the Power of VLMs in Autonomous Driving via Reinforcement Learning and Reasoning	2025	VLM Reasoning · Reinforcement Learning · Long-Tail Driving	💻

2024

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
EMMA EMMA: End-to-End Multimodal Model for Autonomous Driving	2024	Multimodal Foundation Model · Trajectory Generation · Joint Driving Tasks	🌐
DriVLMe DriVLMe: Enhancing LLM-Based Autonomous Driving Agents with Embodied and Social Experiences	IROS 2024	Embodied Experience · Social Reasoning · Driving Agents	💻
DriveVLM DriveVLM: The Convergence of Autonomous Driving and Large Vision-Language Models	CoRL 2024	Vision-Language Models · Hierarchical Planning · Hybrid Driving System	💻 🌐 The official GitHub repository hosts the paper and project website; model implementation is not released.
VLP VLP: Vision Language Planning for Autonomous Driving	CVPR 2024	Vision-Language Planning · Long-Tail Reasoning · Motion Planning
DriveLM DriveLM: Driving with Graph Visual Question Answering	ECCV 2024	Graph VQA · Driving Reasoning · End-to-End Planning	💻 🌐
LMDrive LMDrive: Closed-Loop End-to-End Driving with Large Language Models	CVPR 2024	Language-Guided Driving · Closed-Loop Control · Instruction Following	💻
Dolphins Dolphins: Multimodal Language Model for Driving	ECCV 2024	Vision-Language Model · Driving Instruction Tuning · In-Context Learning	💻 🌐 Partial code: Inference code and checkpoint are public; training code and benchmark release remain pending.
On the Road with GPT-4V On the Road with GPT-4V(ision): Early Explorations of Visual-Language Model on Autonomous Driving	ICLR 2024	Vision-Language Models · Driving Reasoning · Open-World Scenarios	💻 Official test images, prompts, and responses are public; this repository is an exploration archive rather than a released driving implementation.
LLM-Driver Driving with LLMs: Fusing Object-Level Vector Modality for Explainable Autonomous Driving	ICRA 2024	Large Language Models · Vectorized Scene Inputs · Explainable Driving	💻
DriveGPT4 DriveGPT4: Interpretable End-to-End Autonomous Driving via Large Language Model	IEEE RA-L 2024	Multimodal LLM · End-to-End Control · Driving Explanations	💻 🌐
DiLu DiLu: A Knowledge-Driven Approach to Autonomous Driving with Large Language Models	ICLR 2024	Large Language Models · Knowledge and Memory · Self-Evolving Driving	💻 🌐

2023

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
DriveMLM DriveMLM: Aligning Multi-Modal Large Language Models with Behavioral Planning States for Autonomous Driving	2023	Multimodal LLM · Behavior Planning · Closed-Loop Driving	💻 Code pending: The official README lists code, models, dataset, and annotations as unreleased.
LanguageMPC LanguageMPC: Large Language Models as Decision Makers for Autonomous Driving	2023	Large Language Models · Motion Planning · Decision Making	🌐
GPT-Driver GPT-Driver: Learning to Drive with GPT	NeurIPS 2023 Workshop	Large Language Models · Trajectory Planning · Numerical Reasoning	💻 🌐
Drive as You Speak Drive as You Speak: Enabling Human-Like Interaction with Large Language Models in Autonomous Vehicles	2023	Large Language Models · Human-Vehicle Interaction · Decision Making

2022

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
DOROTHIE DOROTHIE: Spoken Dialogue for Handling Unexpected Situations in Interactive Autonomous Driving Agents	Findings of EMNLP 2022	Spoken Dialogue · Interactive Driving · Human-Agent Collaboration	💻 🌐

Driving World Models

74 papers; 42 with a verified public code link.

2026

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
TerrainForge TerrainForge: Physics-Grounded road geometry Editing for Counterfactual Autonomous Driving	2026	Counterfactual Simulation · Road Geometry Editing · Vehicle Dynamics
AD-E2E-JEPA AD-E2E-JEPA: A Joint-Embedding Predictive Architecture For End-to-End Autonomous Driving	2026	Joint-Embedding Prediction · Goal-Conditioned Planning · Efficient World Model	💻 Code pending: The author announces source-code release by October 19, 2026; the current repository contains a README and figures.
DreamStream DreamStream: Towards Policy-Oriented Generative Simulation for End-to-End Driving	CoRL 2026	Generative Simulation · Closed-Loop Evaluation · Policy-Oriented Fidelity	💻 🌐 Code pending
ForeDrive ForeDrive: Foresight-Guided End-to-End Autonomous Driving with a Planning-Relevant Latent World Model	2026	Latent World Model · JEPA · Future-Conditioned Planning
ZYT-World ZYT-World: A Real-Time Controllable World Model for Closed-Loop Autonomous-Driving Simulation	2026	Driving World Model · Controllable Simulation
MM-Future MM-Future: Multi-Mode Joint World-Action Modeling for Autonomous Driving	2026	Multimodal Future Prediction · World Modeling
SV-WAM SV-WAM: An Efficient Surround-View World-Action Model for End-to-End Autonomous Driving	2026	World-Action Model · Semantic Video Representation
VATIX How Far Can 5,500 Hours of Driving Take You? A Scaling Law Analysis of Video Diffusion Models	2026	Video Diffusion · Driving Data Scaling · World Generation	💻
BehaviorWorldGen BehaviorWorldGen: Closing the Loop between Action Models and World Simulators via Controllable Behavior-Aware Structured World Generation	2026	Behavior-Conditioned Generation · Driving Simulation · World Model
WA-JEPA WA-JEPA: Rethinking the Video JEPA Paradigm for World-Action Modeling in Autonomous Driving	2026	Video JEPA · World-Action Modeling · Self-Supervised Learning	💻
RISE RISE: Adaptive Imagination for World Action Models	2026	World-Action Modeling · Action-Conditioned Prediction
DA-WAM DA-WAM: Decision-Aligned Future Latents for Driving World Models	2026	World-Action Modeling · Latent Dynamics
4D-WAM 4D-WAM: 4D Consistent World Modeling for Autonomous Driving	2026	World-Action Modeling · 4D Scene Representation
SimWAM SimWAM: A Simple World Action Model for End-to-End Autonomous Driving	2026	World-Action Model · Training-Time Prediction · End-to-End Driving	💻
SUV SUV: Future Scene Understanding as Video Generation for End-to-End Driving	2026	Video Generation · Scene Understanding · End-to-End Driving	💻 Partial code: Checkpoint and inference code are public; training and the WOD-E2E pipeline remain pending.
Auto-JEPA Auto-JEPA: A Latent World Model of Continuous Intent for End-to-End Autonomous Driving	2026	Action-Centered JEPA · Latent Dynamics · Planning	💻 Code pending: The paper links this repository, which currently contains a scaffold; model components are still being organized.
INTACT INTACT: Isomorphic Intent-to-Action Learning for Search-Free World Models	2026	Intent-to-Action · JEPA · Reward-Free Learning	💻
HyWorldVLA HyWorldVLA: A Vision-Language-Action Model with Hybrid World Modeling for Autonomous Driving	2026	VLA-WAM Integration · Latent World Modeling · Driving Policy
GeoWorldAD GeoWorldAD: Geometry World Action Model for Autonomous Driving	2026	Geometric Grounding · 3D Scene Modeling · Future-Aware Planning	💻 Code pending: The repository links the paper; training and inference code are still listed as coming soon.
OWMDrive OWMDrive: Causality-Aware End-to-End Autonomous Driving via 4D Occupancy World Model	2026	Temporal Causal Dynamics · World Modeling · End-to-End Planning
Metis Metis: A Generalizable and Efficient World-Action Model for Autonomous Driving and Urban Navigation	2026	World-Action Modeling · Efficient Inference · Decoupled Video and Action	💻 Code pending: The repository contains the project description and assets; inference and training code remain pending.
NVIDIA OmniDreams NVIDIA OmniDreams: Real-Time Generative World Model for Closed-Loop Autonomous Vehicle Simulation	2026	Interactive Simulation · Closed-Loop Evaluation · Long-Tail Scenarios	💻
Cosmos 3 Cosmos 3: Omnimodal World Models for Physical AI	2026	Omnimodal Foundation Model · World Generation · Physical AI	💻 🌐
X-Foresight X-Foresight: A Joint Vision-Action Causal Forecasting Network via Predictive World Modeling	2026	Video-Based World Modeling · Future Reasoning · VLA
Xiaomi EV World Model Xiaomi Auto World Model: A Joint World Model Integrating Reconstruction and Generation for Autonomous Driving	2026	3D Reconstruction · World Generation · Autonomous Driving
GEM GEM: Gaussian Evolution Model for Occupancy Forecasting and Motion Planning	2026	Occupancy Forecasting · 3D Scene Dynamics · Motion Planning
EponaV2 EponaV2: Driving World Model with Comprehensive Future Reasoning	2026	World Model · Data-Efficient Planning · Annotation-Free Learning	💻 Code pending: The repository currently contains only a README and license; no implementation is present.
HorizonDrive HorizonDrive: Self-Corrective Autoregressive World Model for Long-horizon Driving Simulation	2026	Autoregressive Simulation · Long-Horizon Rollout · Self-Correction	💻
DAWN The DAWN of World-Action Interactive Models	2026	World-Action Co-Modeling · Joint Prediction and Planning	💻
DeepSight DeepSight: Long-Horizon World Modeling via Latent States Prediction for End-to-End Autonomous Driving	2026	Latent-State Prediction · Long-Horizon World Modeling · VLM Reasoning	💻
CoWorld-VLA CoWorld-VLA: Thinking in a Multi-Expert World Model for Autonomous Driving	2026	Latent World Reasoning · Planning-Oriented VLA · 3D Grounding	💻
DriveFuture DriveFuture: Future-Aware Latent World Models for Autonomous Driving	2026	Latent World Model · Future-Conditioned Planning	🌐
X-Cache X-Cache: Cross-Chunk Block Caching for Few-Step Autoregressive World Models Inference	2026	Real-Time Simulation · Autoregressive Video Generation · Inference Efficiency
DriveVA DriveVA: Video Action Models are Zero-Shot Drivers	ECCV 2026	World-Model Planning · Domain Generalization · Robustness	💻
ExploreVLA ExploreVLA: Dense World Modeling and Exploration for End-to-End Autonomous Driving	ECCV 2026	Policy Exploration · Reinforcement Learning · World Modeling	💻 🌐 Partial code: Data preprocessing and configuration files are public; model/training code and checkpoints remain pending.
DriveDreamer-Policy DriveDreamer-Policy: A Geometry-Grounded World-Action Model for Unified Generation and Planning	2026	Geometry-Grounded World Model · Unified Generation and Planning	💻 🌐
Uni-World VLA Uni-World VLA: Interleaved World Modeling and Planning for Autonomous Driving	ECCV 2026	Unified VLA and World Model · Closed-Loop Planning	💻
Latent-WAM Latent-WAM: Latent World Action Modeling for End-to-End Autonomous Driving	2026	Latent World-Action Model · Spatial Representation · Temporal Dynamics
X-World X-World: Controllable Ego-Centric Multi-Camera World Models for Scalable End-to-End Driving	2026	Simulation Benchmark · VLA Evaluation · Controllable Scenarios
Fast-WAM Fast-WAM: Do World Action Models Need Test-time Future Imagination?	2026	World-Action Model · Training-Time Imagination · Inference Efficiency	💻 🌐
WorldDrive Bridging Scene Generation and Planning: Driving with World Model via Unifying Vision and Motion Representation	2026	WorldDrive · Scene Generation · Motion Planning	💻
ResWorld ResWorld: Temporal Residual World Model for End-to-End Autonomous Driving	ICLR 2026	Temporal Residual Dynamics · World-Model Planning	💻
DriveWorld-VLA DriveWorld-VLA: Unified Latent-Space World Modeling with Vision-Language-Action for Autonomous Driving	2026	Unified VLA and World Model · Latent Dynamics · Joint Planning	💻
UniDWM UniDWM: Towards a Unified Driving World Model via Multifaceted Representation Learning	2026	Unified Driving World Model · Geometry and Dynamics
Drive-JEPA Drive-JEPA: Video JEPA Meets Multimodal Trajectory Distillation for End-to-End Driving	2026	JEPA · Self-Supervised Video Learning · Planning Representation	💻
UniDrive-WM UniDrive-WM: Unified Understanding, Planning and Generation World Model for Autonomous Driving	2026	Unified VLM Planning · World Modeling · End-to-End Driving	💻 🌐 The repository hosts the official project website; the Code button does not link to a released implementation.

2025

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
DriveLaW DriveLaW:Unifying Planning and Video Generation in a Latent Driving World	2025	World-Action Model · Video Generation · Joint Planning	💻 🌐 Official repository transferred from wm-research/DriveLaW to Xiaomi Research.
WorldRFT WorldRFT: Latent World Model Planning with Reinforcement Fine-Tuning for Autonomous Driving	2025	Reinforcement Fine-Tuning · Latent World Model · End-to-End Planning	💻 Code pending: The author repository currently contains a README and license; the README says code will be uploaded later.
MindDrive MindDrive: An All-in-One Framework Bridging World Models and Vision-Language Model for End-to-End Autonomous Driving	2025	Trajectory Scoring · World-Model Planning · Multi-Modal Planning	💻 🌐
SimScale SimScale: Learning to Drive via Real-World Simulation at Scale	2025	Driving Simulation · Synthetic Data Generation · Policy Learning	💻 🌐
DriveVLA-W0 DriveVLA-W0: World Models Amplify Data Scaling Law in Autonomous Driving	2025	World-Model Pretraining · Data Scaling Laws · Policy Learning	💻
WorldMirror WorldMirror: Universal 3D World Reconstruction with Any-Prior Prompting	2025	3D Reconstruction · Geometric Foundation Model · World Representation	💻
RAP RAP: 3D Rasterization Augmented End-to-End Planning	2025	Alternative-Trajectory Generation · Closed-Loop Recovery · World Modeling	💻 🌐
IRL-VLA IRL-VLA: Training an Vision-Language-Action Policy via Reward World Model	2025	Inverse Reinforcement Learning · Policy Post-Training · VLA	💻 Code pending: The repository contains the paper description and figures; no implementation is present.
World4Drive World4Drive: End-to-End Autonomous Driving via Intention-aware Physical Latent World Model	ICCV 2025	Self-Supervised World Model · Annotation-Free Planning	💻
EponaV1 Epona: Autoregressive Diffusion World Model for Autonomous Driving	ICCV 2025	Autoregressive Video Diffusion · Long-Horizon Prediction · Planning	💻 🌐
WoTE End-to-End Driving with Online Trajectory Evaluation via BEV World Model	2025	Online Trajectory Evaluation · BEV World Model · Closed-Loop Planning	💻
Cosmos-Transfer1 Cosmos-Transfer1: Conditional World Generation with Adaptive Multimodal Control	2025	Controllable World Generation · Multi-Modal Conditioning · Simulation	💻
HERMES HERMES: A Unified Self-Driving World Model for Simultaneous 3D Scene Understanding and Generation	ICCV 2025	3D Scene Understanding · World Generation · Unified World Model	💻
Cosmos Cosmos World Foundation Model Platform for Physical AI	2025	World Foundation Model · Video Generation · Physical AI	💻

2024

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
GaussianAD GaussianAD: Gaussian-Centric End-to-End Autonomous Driving	2024	Gaussian Representation · 3D Scene Modeling · End-to-End Planning	💻
LAW Enhancing End-to-End Autonomous Driving with Latent World Model	ICLR 2025	Latent World Model · Self-Supervised Representation · End-to-End Planning	💻
Vista Vista: A Generalizable Driving World Model with High Fidelity and Versatile Controllability	NEURIPS 2024	Driving World Model · Controllable Generation · Generalization	💻 🌐
GenAD Generalized Predictive Model for Autonomous Driving	CVPR 2024	Driving Video Prediction · Large-Scale Pretraining · Zero-Shot Generalization	💻 🌐 Partial code: OpenDV download, preprocessing, and annotation tools are public; a GenAD predictive-model implementation was not verified.
WoVoGen WoVoGen: World Volume-aware Diffusion for Controllable Multi-camera Driving Scene Generation	ECCV 2024	World Volume · Multi-Camera Generation · Diffusion Model	💻
Drive-WM Driving into the Future: Multiview Visual Forecasting and Planning with World Model for Autonomous Driving	CVPR 2024	Multi-View Forecasting · Counterfactual Planning · World Model	💻 🌐 Code pending: The repository contains the Diffusers codebase; paper-specific model weights and training code remain pending.
DrivingDiffusion DrivingDiffusion: Layout-Guided multi-view driving scene video generation with latent diffusion model	ECCV 2024	Latent Diffusion · Multi-View Video · Layout Control	💻 Code pending: Some configurations and custom Diffusers components are present; the README still lists training and inference as coming soon.
DriveDreamer DriveDreamer: Towards Real-world-driven World Models for Autonomous Driving	ECCV 2024	Diffusion World Model · Real-World Driving · Controllable Generation	💻

2023

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
OccWorld OccWorld: Learning a 3D Occupancy World Model for Autonomous Driving	2023	Occupancy World Model · 3D Scene Forecasting · Planning	💻
ADriver-I ADriver-I: A General World Model for Autonomous Driving	2023	Vision-Action Modeling · Autoregressive Prediction · Driving Control
GAIA-1 GAIA-1: A Generative World Model for Autonomous Driving	2023	Generative World Model · Video Prediction · Action Conditioning

2022

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
MILE Model-Based Imitation Learning for Urban Driving	NeurIPS 2022	Latent Dynamics · Model-Based Imitation · Urban Driving	💻

2021

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
World on Rails Learning to Drive from a World on Rails	ICCV 2021	Factorized World Model · Model-Based Imitation · CARLA	💻 🌐
DriveGAN DriveGAN: Towards a Controllable High-Quality Neural Simulation	CVPR 2021	Controllable Simulation · Action-Conditioned Generation · Neural World Model	💻

Recursive Self-Improvement for Autonomous Driving

18 papers; 6 with a verified public code link.

2026

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
RoXDrive RoXDrive: Closed-Loop Reinforcement Learning for End-to-End Autonomous Driving via Action-Faithful Rollouts	2026	World-Model Feedback · Closed-Loop Policy Improvement · Action-Faithful Rollouts	💻 🌐 Code pending: The official repository contains a README and project-page assets; implementation is not released.
OPTED OPTED: On-Policy Fine-Tuning for End-to-End Driving using a Render-Free Teacher	2026	On-Policy Fine-Tuning · Privileged Teacher · Reinforcement Learning	🌐
DriveZero DriveZero: End-to-End Driving Beyond Human Demonstrations	2026	Self-Play · Closed-Loop Reinforcement Learning · Teacher Distillation	💻 🌐 Partial code: DriveRL inference code and checkpoints are public; DriveZero and DriveVFM remain unchecked on the release roadmap.
Self-Improving Autonomous Vehicles Self-Improving Autonomous Vehicles via Real-World Reinforcement Learning	IJCAI 2026	Real-World Reinforcement Learning · Informative Scenario Collection · Safety-Aware Exploration
CL4AD Scaling Curriculum Learning For Autonomous Driving	Under review at NeurIPS 2026	Curriculum Learning · Reinforcement Learning · Sample Efficiency	💻
AlignADV From Attacks to Curricula: Learnability-Guided Adversarial Training for Safe Autonomous Driving	2026	Adversarial Training · Curriculum Learning · Safety	🌐
EvoDrive EvoDrive: Pareto Evolution for Safety-Critical Autonomous Driving via Self-Improving LLM Agents	arXiv 2026	Agentic Evolution · Safety-Critical Scenario Generation · Pareto Optimization	💻 🌐 The GitHub repository hosts the project website only; no implementation was verified.
Agentic AI + SkillBank Agentic AI in Autonomous Driving: LLM-Enhanced RL with Reusable Skills and Adaptive Exploration for Enhanced Safety	ICAART 2026	LLM-Guided Reinforcement Learning · Reusable Driving Skills · Continual Self-Improvement
Dense Learning Breaking through safety performance stagnation in autonomous vehicles with dense learning	Nature Communications 2026	Dense Reinforcement Learning · Safety-Critical Learning · Real-World Testing	💻

2025

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
Safe-Evolving RL An Uncertainty-Aware Safe-Evolving Reinforcement Learning Algorithm for Decision-Making and Control in Highway Autonomous Driving	Eng. Appl. Artif. Intell. 2025	Safe Online Reinforcement Learning · Policy Improvement Stability · Uncertainty-Aware Control
Mechanism-Experience Learning A Safe and Efficient Self-Evolving Algorithm for Decision-Making and Control of Autonomous Driving Systems	IEEE T-ITS 2025	Self-Evolving Driving Policy · Mechanism-Experience Learning · Safety-Constrained Optimization
Expert-Agent Feedback Integrating Expert-Agent Bi-Directional Positive Feedback for Autonomous Driving Through Reinforcement Learning	Intelligent Transportation Engineering 2025	Expert-Agent Co-Evolution · Online Interaction Data · Reinforcement Learning	💻
Safe Self-Evolution with Risk Quantification A Safe Self-Evolution Algorithm for Autonomous Driving Based on Data-Driven Risk Quantification Model	Accid. Anal. Prev. 2025	Risk-Aware Self-Evolution · Safe Driving Policy · Simulation and Real-Vehicle Testing
Gigaflow Robust Autonomy Emerges from Self-Play	ICML 2025	Self-Play · Closed-Loop Reinforcement Learning · Large-Scale Simulation	🌐

2024

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
Human Feedback for Intelligent Driving Human Feedback Enhanced Autonomous Intelligent Systems: A Perspective from Intelligent Driving	Autonomous Intelligent Systems 2024	Human Feedback · Self-Evolving Intelligent Driving · Interactive Reinforcement Learning

2023

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
DCARL Continuous Improvement of Self-Driving Cars Using Dynamic Confidence-Aware Reinforcement Learning	Nat. Mach. Intell. 2023	Confidence-Aware Reinforcement Learning · Continuous Policy Improvement · Real-World Driving	💻

2022

🧠 Method	🗓️ Year / Venue	🏷️ Tags	🔗 Resources
Self-Improving Safety RL Self-Improving Safety Performance of Reinforcement Learning Based Driving with Black-Box Verification Algorithms	arXiv 2022	Black-Box Verification · Failure-Driven Policy Updates · Safety-Critical Scenarios	💻
Trustworthy Improvement RL Trustworthy Safety Improvement for Autonomous Driving Using Reinforcement Learning	Transp. Res. Part C 2022	Trustworthy Reinforcement Learning · Baseline-Guarded Improvement · Highway Safety

Public Datasets

Dataset access terms and releases follow the original providers. The source and inspiration repositories are listed in Acknowledgements.

Perception and Multimodal Understanding

29 datasets.

Dataset / Method and Paper Title	Task	Scale	Access	Resources
KITTI The KITTI Vision Benchmark Suite	Perception, Tracking, and Depth	Stereo video; LiDAR; Karlsruhe routes	Open	🌍 📄
Cityscapes The Cityscapes Dataset for Semantic Urban Scene Understanding	Semantic Segmentation	5,000 finely annotated images; 50 cities	Registration	🌍 📄
nuScenes A Multimodal Dataset for Autonomous Driving	3D Detection and Tracking	1,000 scenes; 6 cameras; LiDAR and radar	Registration	🌍 📄 💻
Waymo Open Dataset Waymo Open Dataset: An Autonomous Driving Dataset	Perception, Motion, and End-to-End Driving	Perception, Motion, and End-to-End subsets	Registration	🌍 📄 💻
Argoverse 2 Argoverse 2: Next Generation Datasets for Self-Driving Perception and Forecasting	Perception and Motion Forecasting	6 US cities; multiple sensor and motion datasets	Open	🌍 📄 💻
PandaSet PandaSet: Advanced Sensor Suite Dataset for Autonomous Driving	3D Detection and Segmentation	100+ scenes; 7 cameras; 2 lidars	Registration	🌍 💻
ONCE One Million Scenes for Autonomous Driving	3D Detection and Self-Supervised Learning	1M scenes; 7 cameras; 1 lidar	Open	🌍 📄 💻
Audi A2D2 A2D2: Audi Autonomous Driving Dataset	Perception and Semantic Segmentation	Around 40,000 frames; 6 cameras; 5 lidars	Open	🌍 💻
BDD100K BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning	Video, Detection, and Segmentation	100K videos; 10 tasks	Registration	🌍 📄 💻
ApolloScape The ApolloScape Open Dataset for Autonomous Driving and Its Application	Perception and Scene Parsing	Multiple cities; stereo and surround-view tasks	Registration	🌍 📄
Honda H3D The H3D Dataset for Full-Surround 3D Multi-Object Detection and Tracking in Crowded Urban Scenes	3D Detection and Tracking	160 scenes; 27,721 frames; 1M labeled instances	Open	🌍 📄
Mapillary Vistas The Mapillary Vistas Dataset for Semantic Understanding of Street Scenes	Semantic and Instance Segmentation	25K images; 66 object categories	Registration	🌍 📄 💻
Talk2Car Taking Control of Your Self-Driving Car	Language Grounding	11,959 commands; 850 videos	Open	🌍 📄 💻
SUTD-TrafficQA A Question Answering Benchmark and an Efficient Network for Video Reasoning over Traffic Events	Traffic Video QA	10,080 videos; 62,535 QA pairs	Request	🌍 📄 💻
DRAMA Joint Risk Localization and Captioning in Driving	Risk Localization & Captioning	17,785 scenarios	Open	🌍 📄
NuScenes-QA A Multi-modal Visual Question Answering Benchmark for Autonomous Driving Scenario	Multimodal Driving QA	34,149 scenes; 459,941 QA pairs	Open	🌍 📄 💻
DriveGPT4 / BDD-X Interpretable End-to-end Autonomous Driving via Large Language Model	Instruction Tuning & Control	16,803 train; 2,123 test clips	Project	🌍 📄
LingoQA Visual Question Answering for Autonomous Driving	Driving Video QA	28K scenarios; 419K annotations	Open	🌍 📄 💻
DriveLM Driving with Graph Visual Question Answering	Graph Visual QA	4,072 train; 799 validation frames	Open	🌍 📄 💻
MAPLM A Real-World Large-Scale Vision-Language Dataset for Map and Traffic Scene Understanding	Map & Traffic Understanding	2M point clouds; panoramic imagery	Open	🌍 💻
NuInstruct Holistic Autonomous Driving Understanding by Bird's-Eye-View Injected Multi-Modal Large Models	Multi-view Instruction QA	91K pairs; 17 subtasks	Open	🌍 📄 💻
CODA-LM Automated Evaluation of Large Vision-Language Models on Self-driving Corner Cases	Corner-case Evaluation	9,768 scenarios; 63K+ annotations	Open	🌍 📄 💻
OmniDrive-nuScenes A Holistic LLM-Agent Framework for Autonomous Driving with 3D Perception, Reasoning and Planning	3D Perception & Reasoning	nuScenes-derived; offline and online QA	Open	🌍 📄 💻
Senna Planning QA Senna: Bridging Large Vision-Language Models and End-to-End Autonomous Driving	Planning-oriented QA	Generated from nuScenes	Generation Scripts	🌍 📄 💻
Rank2Tell A Multimodal Driving Dataset for Joint Importance Ranking and Reasoning	Importance Ranking and Reasoning	Driving videos with dense QA annotations	Open	🌍 📄
doScenes An Autonomous Driving Dataset with Natural Language Instruction for Human Interaction and Vision-Language Navigation	Natural-Language Driving Instructions	Public instruction and referentiality annotations	Open	🌍 📄 💻
nuScenes-MQA Integrated Evaluation of Captions and QA for Autonomous Driving Datasets using Markup Annotations	Captioning and Visual QA	nuScenes-based multimodal captions and QA	Research Release	🌍 📄
BDD-OIA Explainable Object-induced Action Decision for Autonomous Vehicles	Explainable Driving Decisions	BDD100K-derived video annotations	Open	🌍 📄
HAD Advice Grounding Human-to-Vehicle Advice for Self-Driving Vehicles	Language Grounding and Advice	Human-to-vehicle advice annotations	Open	🌍 📄

Motion, Planning, and Cooperative Driving

10 datasets.

Dataset / Method and Paper Title	Task	Scale	Access	Resources
INTERACTION The INTERACTION Dataset: A Road Intersection-Centric Autonomous Driving Dataset	Motion Forecasting and Planning	11 locations; 10 Hz trajectories	Open	🌍 📄 💻
highD The highD Dataset: A Drone Dataset of Naturalistic Vehicle Trajectories on German Highways	Vehicle Trajectory Prediction	110,000 vehicles; 60 recording hours	Request	🌍 📄
Waymo Open Motion Dataset Large Scale Interactive Motion Forecasting for Autonomous Driving	Interactive Motion Forecasting	More than 100,000 20-second segments	Registration	🌍 📄 💻
nuPlan nuPlan: A Closed-Loop ML-Based Planning Benchmark for Autonomous Vehicles	Planning and Closed-Loop Simulation	1,282 hours; 4 cities	Registration	🌍 📄 💻
PIE PIE: A Large-Scale Dataset and Models for Pedestrian Intention Estimation and Trajectory Prediction	Pedestrian Intention and Trajectory Prediction	6+ hours; 300K+ labeled frames; 1,842 pedestrians	Open	🌍 📄 💻
Lyft Level 5 One Thousand and One Hours: Self-driving Motion Prediction Dataset	Motion Forecasting and Planning	1,118 hours; about 170K scenes	Request	🌍 📄 💻
DAIR-V2X DAIR-V2X: A Large-Scale Benchmark for Vehicle-Infrastructure Cooperative 3D Object Detection	Vehicle-Infrastructure Cooperative Perception	Vehicle and infrastructure sensor pairs	Open	🌍 📄 💻
V2X-Seq V2X-Seq: A Large-Scale Sequential Dataset for Vehicle-Infrastructure Cooperative Perception and Forecasting	Cooperative Perception and Forecasting	4 scenarios; sequential multi-agent data	Open	🌍 📄 💻
V2V4Real V2V4Real: A Real-world Large-scale Dataset for Vehicle-to-Vehicle Cooperative Perception	Vehicle-to-Vehicle Cooperative Perception	410 km; 20K LiDAR frames; 40K RGB frames; 240K boxes	Open	🌍 📄 💻
OPV2V OPV2V: An Open Benchmark Dataset and Fusion Pipeline for Perception with Vehicle-to-Vehicle Communication	Cooperative Perception	73 scenes; 11,464 frames	Open	🌍 📄 💻

Simulation and Synthetic Data

7 datasets.

Dataset / Method and Paper Title	Task	Scale	Access	Resources
CARLA Simulator CARLA: An Open Urban Driving Simulator	Closed-Loop Driving Simulation	Open-source simulator; multiple towns and scenarios	Open	🌍 📄 💻
Virtual KITTI 2 Virtual KITTI 2	Synthetic Perception and Domain Transfer	5 scenarios; 5 weather and imaging conditions	Open	🌍 📄
SHIFT SHIFT: A Synthetic Driving Dataset for Continuous Multi-Task Domain Adaptation	Synthetic Multi-Task Perception	70+ hours; 2.5M annotated frames	Open	🌍 📄 💻
DeepAccident DeepAccident: A Motion and Accident Prediction Benchmark for V2X Autonomous Driving	Accident and Cooperative Perception	57K annotated frames; 285K samples; 691 scenarios	Open	🌍 📄 💻
NAVSIM Data-Driven Non-Reactive Autonomous Vehicle Simulation and Benchmarking	Planning Evaluation	OpenScene-derived; navtrain and navtest splits	Open	🌍 📄 💻
Bench2Drive Towards Multi-Ability Benchmarking of Closed-Loop End-to-End Autonomous Driving	Interactive Closed-Loop Driving	220 routes; 2M+ frames	Open	🌍 📄 💻
HUGSIM HUGSIM: A Real-Time, Photo-Realistic and Closed-Loop Simulator for Autonomous Driving	Photorealistic Closed-Loop Simulation	Real-world reconstructed driving scenes	Open	🌍 📄 💻

Public Evaluation Benchmarks

Benchmarks are grouped by evaluation protocol and research task; scores from different protocols remain separate.

Closed-Loop Planning

3 benchmarks.

Benchmark	Track	Evaluation Setting	Primary Metric	Resources
Bench2Drive	End-to-End Planning	CARLA closed-loop	Driving Score	🌍 📄 💻
nuPlan	Planning	Reactive closed-loop	Closed-loop score	🌍 📄 💻
CARLA Leaderboard 2.0	End-to-End Planning	Interactive closed-loop	Driving Score	🌍 💻

Open-Loop and Non-Reactive Planning

2 benchmarks.

Benchmark	Track	Evaluation Setting	Primary Metric	Resources
NAVSIM	End-to-End Planning	Non-reactive simulation	PDMS	🌍 📄 💻
nuScenes Planning	Planning / World Model	Open-loop log replay	Avg. L2 / Collision	🌍 📄 💻

VLA and Driving Reasoning

2 benchmarks.

Benchmark	Track	Evaluation Setting	Primary Metric	Resources
DriveLM	VLA / Reasoning	Graph VQA	Accuracy / language metrics	🌍 📄 💻
Reason2Drive	VLA / Reasoning	Video-text reasoning	Reasoning score	🌍 📄 💻

Reliability and Robustness

1 benchmark.

Benchmark	Track	Evaluation Setting	Primary Metric	Resources
DriveBench	VLA / Reliability	17 visual reliability settings	Accuracy / language / GPT score	🌍 📄 💻

Leaderboard

Benchmark results are stored in data/leaderboard.json. Each entry preserves its evaluation protocol, data split, and reproducibility evidence. Results from incompatible protocols are not ranked together.

License

This project is released under the Apache License 2.0.

Acknowledgements

We thank the maintainers of GE2EAD, GenAI4AD, and ReCogDrive for their research collections and dataset references, which helped inform this catalog. Please consult the original projects for their own licenses and access terms.