:sunglasses: Awesome VLA for Autonomous Driving
Autonomous driving has long relied on modular "Perception-Decision-Action" pipelines, whose hand-crafted interfaces and rule-based components often struggle in complex, dynamic, or long-tailed scenarios. Their cascaded structure also amplifies upstream perception errors, undermining downstream planning and control.
This survey reviews vision-action (VA) models and vision-language-action (VLA) models for autonomous driving. We trace the evolution from early VA approaches to modern VLA frameworks, and organize existing methods into two principal paradigms:
- End-to-End VLA, which integrates perception, reasoning, and planning within a single model.
- Dual-System VLA, which separates slow deliberation (via VLMs) from fast, safety-critical execution (via planners).
 |
|---|
For more details, kindly refer to our :books: Paper, :globe_with_meridians: Project Page, and :hugs: HuggingFace Leaderboard.
:books: Citation
If you find this work helpful for your research, please kindly consider citing our paper:
@article{survey_vla4ad,
title = {Vision-Language-Action Models for Autonomous Driving: Past, Present, and Future},
author = {Tianshuai Hu and Xiaolu Liu and Song Wang and Yiyao Zhu and Ao Liang and Lingdong Kong and Guoyang Zhao and Zeying Gong and Jun Cen and Zhiyu Huang and Xiaoshuai Hao and Linfeng Li and Hang Song and Xiangtai Li and Jun Ma and Shaojie Shen and Jianke Zhu and Dacheng Tao and Ziwei Liu and Junwei Liang},
journal = {arXiv preprint arXiv:2512.16760},
year = {2025},
}
@article{survey_3d_4d_world_models,
title = {{3D} and {4D} World Modeling: A Survey},
author = {Lingdong Kong and Wesley Yang and Jianbiao Mei and Youquan Liu and Ao Liang and Dekai Zhu and Dongyue Lu and Wei Yin and Xiaotao Hu and Mingkai Jia and Junyuan Deng and Kaiwen Zhang and Yang Wu and Tianyi Yan and Shenyuan Gao and Song Wang and Linfeng Li and Liang Pan and Yong Liu and Jianke Zhu and Wei Tsang Ooi and Steven C. H. Hoi and Ziwei Liu},
journal = {arXiv preprint arXiv:2509.07996},
year = {2025}
}
Table of Contents
- 1. Vision-Action Models
- 2. Vision-Language-Action Models
- 3. Datasets & Benchmarks
- 4. Applications
- 5. Other Resources
1. Vision-Action Models
:one: Action-Only Models
:timer_clock: In chronological order, from the earliest to the latest.
| Model | Paper | Venue | Website | GitHub |
|---|---|---|---|---|
Selective Memory |
||||
| Not All History Helps: Velocity-Aware Selective Memory for Long-Horizon End-to-End Autonomous Driving | arXiv 2026 | - | - | |
MomADv2 |
||||
| MomADv2: Reliable Temporal Memory for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
RoG-DAgger |
||||
| RoG-DAgger: Rollout-Guided Post-Training for End-to-End Driving | arXiv 2026 | - | - | |
CL4AD |
||||
| Scaling Curriculum Learning For Autonomous Driving | arXiv 2026 | - | - | |
VG-SAF |
||||
| Variance-Guided Spatial Attention Fusion for Robust End-to-End Driving under Asymmetric Sensor Degradation | arXiv 2026 | - | - | |
LBC |
||||
| Learning by Cheating | CoRL 2020 | - | ||
Latent-DRL |
||||
| End-to-End Model-Free Reinforcement Learning for Urban Driving using Implicit Affordances | CVPR 2020 | - | - | |
NEAT |
||||
| NEAT: Neural Attention Fields for End-to-End Autonomous Driving | ICCV 2021 | - | ||
Roach |
||||
| End-to-End Urban Driving by Imitating a Reinforcement Learning Coach | ICCV 2021 | |||
WoR |
||||
| Learning to Drive from A World on Rails | ICCV 2021 | |||
TCP |
||||
| Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baseline | NeurIPS 2022 | - | ||
Urban-Driver |
||||
| Urban Driver: Learning to Drive from Real-world Demonstrations Using Policy Gradients | CoRL 2022 | |||
LAV |
||||
| Learning from All Vehicles | CVPR 2022 | |||
TransFuser |
||||
| TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Driving | TPAMI 2023 | - | ||
GRI |
||||
| GRI: General Reinforced Imitation and its Application to Vision-Based Autonomous Driving | Robotics 2023 | - | - | |
BEVPlanner |
||||
| Is Ego Status All You Need for Open-Loop End-to-End Autonomous Driving? | CVPR 2024 | - | ||
Raw2Drive |
||||
| Raw2Drive: Reinforcement Learning with Aligned World Models for End-to-End Autonomous Driving (in CARLA v2) | NeurIPS 2025 | - | - | |
RAD |
||||
| RAD: Training an End-to-End Driving Policy via Large-Scale 3DGS-based Reinforcement Learning | NeurIPS 2025 | - | ||
TrajDiff |
||||
| TrajDiff: End-to-End Autonomous Driving without Perception Annotation | arXiv 2025 | - | ||
SimScale |
||||
| SimScale: Learning to Drive via Real-World Simulation at Scale | arXiv 2025 | |||
| - | ||||
| Addressing the Waypoint-Action Gap in End-to-End Autonomous Driving via Vehicle Motion Models | arXiv 2026 | - | - | |
Velocity-Aware Selective Memory |
||||
| Not All History Helps: Velocity-Aware Selective Memory for Long-Horizon End-to-End Autonomous Driving | arXiv 2026 | - | - | |
Scaling Curriculum Learning |
||||
| Scaling Curriculum Learning For Autonomous Driving | arXiv 2026 | - | - | |
MomADv2 |
||||
| MomADv2: Reliable Temporal Memory for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
RoG-DAgger |
||||
| RoG-DAgger: Rollout-Guided Post-Training for End-to-End Driving | arXiv 2026 | - | - | |
DriveZero |
||||
| DriveZero: End-to-End Driving Beyond Human Demonstrations | arXiv 2026 | - | - | |
EMPlan |
||||
| Efficient Multi-Modal Planning with Reward-Guided Preference Optimization for Autonomous Driving | arXiv 2026 | - | - | |
RADP |
||||
| Learning to Explain While Planning: Rule-Aligned Diffusion Planning for Autonomous Driving | arXiv 2026 | - | - | |
:two: Perception-Action Models
:timer_clock: In chronological order, from the earliest to the latest.
| Model | Paper | Venue | Website | GitHub |
|---|---|---|---|---|
ST-P3 |
||||
| ST-P3: End-to-End Vision-Based Autonomous Driving via Spatial-Temporal Feature Learning | ECCV 2022 | - | ||
UniAD |
||||
| Planning-Oriented Autonomous Driving | CVPR 2023 | - | ||
VAD |
||||
| VAD: Vectorized Scene Representation for Efficient Autonomous Driving | ICCV 2023 | - | ||
OccNet |
||||
| Scene as Occupancy | ICCV 2023 | - | ||
GenAD |
||||
| GenAD: Generative End-to-End Autonomous Driving | ECCV 2024 | - | ||
PARA-Drive |
||||
| PARA-Drive: Parallelized Architecture for Real-Time Autonomous Driving | CVPR 2024 | - | ||
Hydra-MDP |
||||
| Hydra-MDP: End-to-End Multimodal Planning with Multi-Target Hydra-Distillation | CVPRW 2024 | |||
SparseAD |
||||
| SparseAD: Sparse Query-Centric Paradigm for Efficient End-to-End Autonomous Driving | arXiv 2024 | - | - | |
GaussianAD |
||||
| GaussianAD: Gaussian-Centric End-to-End Autonomous Driving | arXiv 2024 | - | - | |
DriveTransformer |
||||
| DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Driving | ICLR 2025 | - | ||
SparseDrive |
||||
| SparseDrive: End-to-End Autonomous Driving via Sparse Scene Representation | ICRA 2025 | - | ||
DiffusionDrive |
||||
| DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving | CVPR 2025 | - | ||
GoalFlow |
||||
| GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation in End-to-End Autonomous Driving | CVPR 2025 | |||
GuideFlow |
||||
| GuideFlow: Constraint-Guided Flow Matching for Planning in End-to-End Autonomous Driving | arXiv 2025 | - | ||
ETA |
||||
| ETA: Efficiency through Thinking Ahead, A Dual Approach to Self-Driving with Large Models | arXiv 2025 | - | ||
Geo |
||||
| Spatial Retrieval Augmented Autonomous Driving | arXiv 2025 | - | - | |
DiffusionDriveV2 |
||||
| DiffusionDriveV2: Reinforcement Learning-Constrained Truncated Diffusion Modeling in End-to-End Autonomous Driving | arXiv 2025 | - | ||
NaviHydra |
||||
| NaviHydra: Controllable Navigation-Guided End-to-End Autonomous Driving with Hydra Distillation | arXiv 2025 | - | - | |
Mimir |
||||
| Mimir: Hierarchical Goal-Driven Diffusion with Uncertainty Propagation for End-to-End Autonomous Driving | arXiv 2025 | - | ||
FROST-Drive |
||||
| FROST-Drive: Scalable and Efficient End-to-End Driving with a Frozen Vision Encoder | arXiv 2026 | - | - | |
DrivoR |
||||
| Driving on Registers | arXiv 2026 | |||
SPS |
||||
| See Less, Drive Better: Generalizable End-to-End Autonomous Driving via Foundation Models Stochastic Patch Selection | arXiv 2026 | - | - | |
BevAD |
||||
| What Matters for Scalable and Robust Learning in End-to-End Driving Planners? | CVPR 2026 | |||
EgoFSD |
||||
| EgoFSD: Ego-Centric Fully Sparse Paradigm with Uncertainty Denoising and Iterative Refinement for Efficient End-to-End Self-Driving | ICRA 2026 | - | ||
BeyondSight |
||||
| BeyondSight: Object Permanence for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
What's |
||||
| What's Hidden Matters: Identifying Planning-Critical Occluded Agents using Vision-Language Models | arXiv 2026 | - | - | |
Plug |
||||
| Plug-and-Play Traffic Element Awareness for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
Off |
||||
| Off the Rails: Hijacking the Scoring Head in Generative End-to-End Driving Planners with Safety-Violating Adversarial Perturbations | arXiv 2026 | - | - | |
Herding |
||||
| Herding End-to-End Autonomous Driving via Neuro-Symbolic Safety Guards | arXiv 2026 | - | - | |
VG-SAF |
||||
| Variance-Guided Spatial Attention Fusion for Robust End-to-End Driving under Asymmetric Sensor Degradation | arXiv 2026 | - | - | |
ViRA |
||||
| Do Better Visual Representations Always Lead to Better End-to-End Autonomous Driving? | arXiv 2026 | |||
DriveMamba |
||||
| DriveMamba: Task-Centric Scalable State Space Model for Efficient End-to-End Autonomous Driving | ICLR 2026 | - | - | |
FPD-Drive |
||||
| FPD-Drive: Future Policy Distillation for End-to-End Autonomous Driving | arXiv 2026 | |||
ConDenseDrive |
||||
| ConDenseDrive: Knowledge-Dense, Token-Sparse Visual Encoder for End-to-End Autonomous Driving | arXiv 2026 | - | - |
:three: Image-Based World Models
:timer_clock: In chronological order, from the earliest to the latest.
| Model | Paper | Venue | Website | GitHub |
|---|---|---|---|---|
DriveDreamer |
||||
| DriveDreamer: Towards Real-World-Driven World Models for Autonomous Driving | ECCV 2024 | |||
GenAD |
||||
| GenAD: Generalized Predictive Model for Autonomous Driving | CVPR 2024 | - | ||
Drive-WM |
||||
| Driving into the Future: Multiview Visual Forecasting and Planning with World Model for Autonomous Driving | CVPR 2024 | |||
DrivingWorld |
||||
| DrivingWorld: Constructing World Model for Autonomous Driving via Video GPT | arXiv 2024 | |||
Imagine-2-Drive |
||||
| Imagine-2-Drive: Leveraging High-Fidelity World Models via Multi-Modal Diffusion Policies | IROS 2025 | - | ||
DrivingGPT |
||||
| DrivingGPT: Unifying Driving World Modeling and Planning with Multi-Modal Autoregressive Transformers | ICCV 2025 | - | ||
Epona |
||||
| Epona: Autoregressive Diffusion World Model for Autonomous Driving | ICCV 2025 | |||
VaViM |
||||
| VaViM and VaVAM: Autonomous Driving through Video Generative Modeling | arXiv 2025 | |||
UniDrive-WM |
||||
| UniDrive-WM: Unified Understanding, Planning and Generation World Model For Autonomous Driving | arXiv 2026 | - | ||
DwD |
||||
| Driving with DINO: Vision Foundation Features as a Unified Bridge for Sim-to-Real Generation in Autonomous Driving | arXiv 2026 | - | - | |
WorldDrive |
||||
| Bridging Scene Generation and Planning: Driving with World Model via Unifying Vision and Motion Representation | arXiv 2026 | - | ||
OmniDreams |
||||
| NVIDIA OmniDreams: Real-Time Generative World Model for Closed-Loop Autonomous Vehicle Simulation | arXiv 2026 | - | - | |
DriveDreamer-Policy |
||||
| DriveDreamer-Policy: A Geometry-Grounded World-Action Model for Unified Generation and Planning | NeurIPS 2026 | |||
DriveVA |
||||
| DriveVA: Video Action Models are Zero-Shot Drivers | ECCV 2026 | - | ||
Xiaomi EV World Model |
||||
| Xiaomi Auto World Model: A Joint World Model Integrating Reconstruction and Generation for Autonomous Driving | arXiv 2026 | - | - | |
HERMES++ |
||||
| HERMES++: Toward a Unified Driving World Model for 3D Scene Understanding and Generation | arXiv 2026 | - | - | |
LMGenDrive |
||||
| LMGenDrive: Bridging Multimodal Understanding and Generative World Modeling for End-to-End Driving | arXiv 2026 | - | - | |
X-World |
||||
| X-World: Controllable Ego-Centric Multi-Camera World Models for Scalable End-to-End Driving | arXiv 2026 | - | - | |
How |
||||
| How Can Driving World Models Do Counterfactual Prediction? | arXiv 2026 | - | - | |
Orbis |
||||
| Orbis 2: A Hierarchical World Model for Driving | arXiv 2026 | - | - | |
M$^\text{4}$World: |
||||
| M$^\text{4}$World: A Multi-view Multimodal Driving World Model for Interactive Object Manipulation and Minute-long Streaming | arXiv 2026 | - | - | |
Is |
||||
| Is Energy Guidance All You Need? Training-Free Norm Injection for Driving World Models | arXiv 2026 | - | - | |
GeoRoute |
||||
| GeoRoute: Geometry-Aware Hybrid Inference for Traffic Future-Frame Prediction | arXiv 2026 | - | - | |
Cam2Sim |
||||
| Cam2Sim: Neural Scenario Reconstruction for Closed-Loop Autonomous Driving Simulation | arXiv 2026 | - | - | |
RealWeather |
||||
| RealWeather: Realistic and Scene-Faithful Weather Translation with Driving World Models | arXiv 2026 | - | - | |
SUV |
||||
| SUV: Future Scene Understanding as Video Generation for End-to-End Driving | arXiv 2026 | - | - | |
ZYT-World |
||||
| ZYT-World: A Real-Time Controllable World Model for Closed-Loop Autonomous-Driving Simulation | arXiv 2026 | - | - | |
RoXDrive |
||||
| RoXDrive: Closed-Loop Reinforcement Learning for End-to-End Autonomous Driving via Action-Faithful Rollouts | arXiv 2026 | - | - | |
PhysWAM |
||||
| PhysWAM: Physically Consistent World Action Model for Autonomous Driving | arXiv 2026 | - | - | |
SV-WAM |
||||
| SV-WAM: An Efficient Surround-View World-Action Model for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
:four: Occupancy-Based World Models
:timer_clock: In chronological order, from the earliest to the latest.
| Model | Paper | Venue | Website | GitHub |
|---|---|---|---|---|
OccWorld |
||||
| OccWorld: Learning a 3D Occupancy World Model for Autonomous Driving | ECCV 2024 | |||
NeMo |
||||
| Neural Volumetric World Models for Autonomous Driving | ECCV 2024 | - | - | |
OccVAR |
||||
| OCCVAR: Scalable 4D Occupancy Prediction via Next-Scale Prediction | OpenReview 2024 | - | - | |
RenderWorld |
||||
| RenderWorld: World Model with Self-Supervised 3D Label | arXiv 2024 | - | - | |
DFIT-OccWorld |
||||
| An Efficient Occupancy World Model via Decoupled Dynamic Flow and Image-assisted Training | arXiv 2024 | - | - | |
Drive-OccWorld |
||||
| Driving in the Occupancy World: Vision-Centric 4D Occupancy Forecasting and Planning via World Models for Autonomous Driving | AAAI 2025 | |||
T³Former |
||||
| Temporal Triplane Transformers as Occupancy World Models | arXiv 2025 | - | - | |
OmniNWM |
||||
| OmniNWM: Omniscient Driving Navigation World Models | arXiv 2025 | - | ||
AD-R1 |
||||
| AD-R1: Closed-Loop Reinforcement Learning for End-to-End Autonomous Driving with Impartial World Models | arXiv 2025 | - | - | |
SparseOccVLA |
||||
| SparseOccVLA: Bridging Occupancy and Vision-Language Models via Sparse Queries for Unified 4D Scene Understanding and Planning | arXiv 2026 | - | ||
GEM |
||||
| GEM: Gaussian Evolution Model for Occupancy Forecasting and Motion Planning | arXiv 2026 | - | - | |
OWMDrive |
||||
| OWMDrive: Causality-Aware End-to-End Autonomous Driving via 4D Occupancy World Model | arXiv 2026 | - | - | |
GaussianDWM++ |
||||
| GaussianDWM++: Language-Grounded 3D Gaussian Driving World Model for Unified Scene Understanding, Editing, and Multi-Modal Generation | arXiv 2026 | - | - | |
V2X-WAM |
||||
| V2X-WAM: A Cooperative World Action Model for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
InfraWorld |
||||
| InfraWorld: Motion-Aware World Modeling for Roadside 4D Occupancy Forecasting | arXiv 2026 | - | Release pending | |
OccWAM |
||||
| OccWAM: Occupancy World-Action Models with Joint Scene-to-Ego Flow Matching | arXiv 2026 | - | - | |
:five: Latent-Based World Models
:timer_clock: In chronological order, from the earliest to the latest.
| Model | Paper | Venue | Website | GitHub |
|---|---|---|---|---|
DA-WAM |
||||
| DA-WAM: Decision-Aligned Future Latents for Driving World Models | arXiv 2026 | - | - | |
WA-JEPA |
||||
| WA-JEPA: Rethinking the Video JEPA Paradigm for World-Action Modeling in Autonomous Driving | arXiv 2026 | - | - | |
GeoWAM |
||||
| GeoWAM: Visual Geometry World Action Models for Autonomous Driving | arXiv 2026 | - | - | |
RISE |
||||
| RISE: Adaptive Imagination for World Action Models | arXiv 2026 | - | - | |
Covariate-Shift |
||||
| Mitigating Covariate Shift in Imitation Learning for Autonomous Vehicles Using Latent Space Generative World Models | arXiv 2024 | - | - | |
World4Drive |
||||
| World4Drive: End-to-End Autonomous Driving via Intention-aware Physical Latent World Model | ICCV 2025 | - | - | |
WoTE |
||||
| End-to-End Driving with Online Trajectory Evaluation via BEV World Model | ICCV 2025 | - | ||
LAW |
||||
| Enhancing End-to-End Autonomous Driving with Latent World Model | ICLR 2025 | - | ||
SSR |
||||
| Navigation-Guided Sparse Scene Representation for End-to-End Autonomous Driving | ICLR 2025 | - | ||
Echo-Planning |
||||
| Echo Planning for Autonomous Driving: From Current Observations to Future Trajectories and Back | arXiv 2025 | - | - | |
SeerDrive |
||||
| Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution | NeurIPS 2025 | - | ||
Drive-JEPA |
||||
| Drive-JEPA: Video JEPA Meets Multimodal Trajectory Distillation for End-to-End Driving | arXiv 2026 | - | ||
GraphWorld |
||||
| GraphWorld: Long-Horizon Planning with World Models for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
Unified Driving Tokens |
||||
| Unified Driving Tokens: Representation- and Geometry-Guided Discrete Tokenizer for Driving World Models and Planning | arXiv 2026 | - | - | |
EponaV2 |
||||
| EponaV2: Driving World Model with Comprehensive Future Reasoning | arXiv 2026 | - | - | |
IDOL |
||||
| IDOL: Inverse-Dynamics-Guided Future Prediction for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
ExploreVLA |
||||
| ExploreVLA: Dense World Modeling and Exploration for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
Kinematics-Aware LWM |
||||
| Kinematics-Aware Latent World Models for Data-Efficient Autonomous Driving | arXiv 2026 | - | - | |
SparseWorld |
||||
| SparseWorld: Enhancing End-to-End Autonomous Driving via World Models with Sparse Scene Representation | arXiv 2026 | - | - | |
LWM Survey |
||||
| Latent World Models for Automated Driving: A Unified Taxonomy, Evaluation Framework, and Open Challenges | arXiv 2026 | - | - | |
World Models |
||||
| World Models: A Comprehensive Survey of Architectures, Methodologies, Reasoning Paradigms, and Applications | arXiv 2026 | - | - | |
DynVLA |
||||
| DynVLA: Learning World Dynamics for Action Reasoning in Autonomous Driving | arXiv 2026 | - | - | |
DriveWorld-VLA |
||||
| DriveWorld-VLA: Unified Latent-Space World Modeling with Vision-Language-Action for Autonomous Driving | arXiv 2026 | - | - | |
CRISP |
||||
| CRISP: A Spatiotemporal Camera-Radar Backbone for Driving via Forecasting-Based World-Model Pretraining | arXiv 2026 | - | - | |
Geographic |
||||
| Geographic Diversity Beats Data Volume for Cross-Domain Generalization in Zero-Label JEPA Driving World Models | arXiv 2026 | - | - | |
Auto |
||||
| Auto-JEPA: A Latent World Model of Continuous Intent for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
BrainWAM |
||||
| BrainWAM: Action-Space Coordination of Semantic Priors and Predictive Dynamics for Autonomous Driving | arXiv 2026 | - | - | |
X-Mind |
||||
| X-Mind: Efficient Visual Chain-of-Thought via Predictive World Model for End-to-End Driving | arXiv 2026 | - | - | |
WCog |
||||
| WCog-VLA: A Dual-Level World-Cognitive Vision-Language-Action Model for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
HyWorldVLA |
||||
| HyWorldVLA: A Vision-Language-Action Model with Hybrid World Modeling for Autonomous Driving | arXiv 2026 | - | - | |
LWDrive |
||||
| LWDrive: Layer-Wise World-Model-Guided Vision-Language Model Planning for Autonomous Driving | arXiv 2026 | - | - | |
DriveCache |
||||
| DriveCache: Action-Aware Caching for Driving World Model Inference | arXiv 2026 | - | - | |
DA-WAM |
||||
| DA-WAM: Decision-Aligned Future Latents for Driving World Models | arXiv 2026 | - | - | |
RISE |
||||
| RISE: Adaptive Imagination for World Action Models | arXiv 2026 | - | - | |
WA-JEPA |
||||
| WA-JEPA: Rethinking the Video JEPA Paradigm for World-Action Modeling in Autonomous Driving | arXiv 2026 | - | ||
GeoWAM |
||||
| GeoWAM: Visual Geometry World Action Models for Autonomous Driving | arXiv 2026 | - | - | |
Drive-HWM |
||||
| Drive-HWM: Hierarchical World Models for Dynamic-Latent Guided Autonomous Driving | arXiv 2026 | - | - | |
MM-Future |
||||
| MM-Future: Multi-Mode Joint World-Action Modeling for Autonomous Driving | arXiv 2026 | - | - | |
D-JEPA |
||||
| D-JEPA: A Decision-Aligned Latent World Model | arXiv 2026 | - | - | |
ForeDrive |
||||
| ForeDrive: Foresight-Guided End-to-End Autonomous Driving with a Planning-Relevant Latent World Model | arXiv 2026 | - | - | |
WALT |
||||
| WALT: Learning World-Model-Aligned Latent Trajectories for Autonomous Driving | arXiv 2026 | - | - | |
MomWorld |
||||
| MomWorld: Momentum-Aware Latent World Model for Long-Horizon Autonomous Driving | arXiv 2026 | - | - | |
AD-E2E-JEPA |
||||
| AD-E2E-JEPA: A Joint-Embedding Predictive Architecture For End-to-End Autonomous Driving | arXiv 2026 | - | ||
World4Scorer |
||||
| World4Scorer: Outcome-Grounded World Modeling for Autonomous Driving | arXiv 2026 | |||
ReWAM |
||||
| ReWAM: Reciprocal World Action Models for Interactive Autonomous Driving | arXiv 2026 | - | ||
VGGTWorld-VLA |
||||
| VGGTWorld-VLA: Intent-Conditioned 3D World Evolution for Autonomous Driving | arXiv 2026 | - | - | |
PlanWAM |
||||
| PlanWAM: Planning-Shaped Future Representations for End-to-End Autonomous Driving | arXiv 2026 | |||
ManeuverFlow |
||||
| ManeuverFlow: Maneuver Hypothesis Generation with Future Evidence for Trajectory Evaluation and Refinement in End-to-End Autonomous Driving | arXiv 2026 | - | ||
MORE |
||||
| MORE: Learning Motion-Oriented Registers for Driving World Models and Planning | arXiv 2026 | - | ||
| Release pending |
2. Vision-Language-Action Models
:one: Textual Action Generator
:timer_clock: In chronological order, from the earliest to the latest.
| Model | Paper | Venue | Website | GitHub |
|---|---|---|---|---|
GAPL |
||||
| GAPL: Grounded Action-effect Policy Learning for LLM-Based Trajectory Planning | arXiv 2026 | - | - | |
MAO-LLM |
||||
| Multi-Agent Orchestration with the Common-Sense Reasoning Capabilities of LLMs for Autonomous Driving | arXiv 2026 | - | - | |
DriveMLM |
||||
| DriveMLM: Aligning Multi-Modal Large Language Models with Behavioral Planning States for Autonomous Driving | arXiv 2023 | - | ||
RAG-Driver |
||||
| RAG-Driver: Generalisable Driving Explanations with Retrieval-Augmented In-Context Learning in Multi-Modal Large Language Model | RSS 2024 | |||
RDA-Driver |
||||
| Making Large Language Models Better Planners with Reasoning-Decision Alignment | ECCV 2024 | - | - | |
DriveLM |
||||
| DriveLM: Driving with Graph Visual Question Answering | ECCV 2024 | |||
DriveGPT4 |
||||
| DriveGPT4: Interpretable End-to-end Autonomous Driving via Large Language Model | RA-L 2024 | - | ||
DriVLMe |
||||
| DriVLMe: Enhancing LLM-based Autonomous Driving Agents with Embodied and Social Experience | IROS 2024 | |||
LLaDA |
||||
| Driving Everywhere with Large Language Model Policy Adaptation | CVPR 2024 | |||
VLAAD |
||||
| VLAAD: Vision and Language Assistant for Autonomous Driving | WACVW 2024 | - | ||
OccLLaMA |
||||
| OccLLaMA: A Unified Occupancy-Language-Action World Model for Understanding and Generation Tasks in Autonomous Driving | arXiv 2024 | - | ||
Doe-1 |
||||
| Doe-1: Closed-Loop Autonomous Driving with Large World Model | arXiv 2024 | |||
LINGO-2 |
||||
| LINGO-2: Driving with Natural Language | - | - | ||
SafeAuto |
||||
| SafeAuto: Knowledge-Enhanced Safe Autonomous Driving with Multimodal Foundation Models | ICML 2025 | - | ||
OpenEMMA |
||||
| OpenEMMA: Open-Source Multimodal Model for End-to-End Autonomous Driving | WACV 2025 | - | ||
ReasonPlan |
||||
| ReasonPlan: Unified Scene Prediction and Decision Reasoning for Closed-loop Autonomous Driving | CoRL 2025 | - | ||
WKER |
||||
| World Knowledge-Enhanced Reasoning Using Instruction-Guided Interactor in Autonomous Driving | AAAI 2025 | - | - | |
OmniDrive |
||||
| OmniDrive: A Holistic LLM-Agent Framework for Autonomous Driving with 3D Perception, Reasoning and Planning | CVPR 2025 | - | ||
S4-Driver |
||||
| S4-Driver: Scalable Self-Supervised Driving Multimodal Large Language Model with Spatio-Temporal Visual Representation | CVPR 2025 | - | ||
Occ-LLM |
||||
| Occ-LLM: Enhancing Autonomous Driving with Occupancy-BasedLarge Language Models | ICRA 2025 | - | - | |
DriveBench |
||||
| Are VLMs Ready for Autonomous Driving? An Empirical Study from the Reliability, Data, and Metric Perspectives | ICCV 2025 | |||
FutureSightDrive |
||||
| FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving | NeurIPS 2025 | |||
ImpromptuVLA |
||||
| Impromptu VLA: Open Weights and Open Data for Driving Vision-Language-Action Models | NeurIPS 2025 | |||
Sce2DriveX |
||||
| Sce2DriveX: A Generalized MLLM Framework for Scene-to-Drive Learning | RA-L 2025 | - | - | |
EMMA |
||||
| EMMA: End-to-End Multimodal Model for Autonomous Driving | TMLR 2025 | - | ||
DriveAgent-R1 |
||||
| DriveAgent-R1: Advancing VLM-Based Autonomous Driving with Hybrid Thinking and Active Perception | arXiv 2025 | - | - | |
Drive-R1 |
||||
| Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning | arXiv 2025 | - | - | |
FastDriveVLA |
||||
| FastDriveVLA: Efficient End-to-End Driving via Plug-and-Play Reconstruction-Based Token Pruning | arXiv 2025 | - | - | |
WiseAD |
||||
| WiseAD: Knowledge Augmented End-to-End Autonomous Driving with Vision-Language Model | arXiv 2025 | |||
AutoDrive-R² |
||||
| AutoDrive-R²: Incentivizing Reasoning and Self-Reflection Capacity for VLA Model in Autonomous Driving | arXiv 2025 | - | - | |
OmniReason |
||||
| OmniReason: A Temporal-Guided Vision-Language-Action Framework for Autonomous Driving | arXiv 2025 | - | - | |
OpenREAD |
||||
| OpenREAD: Reinforced Open-Ended Reasoning for End-to-End Autonomous Driving with LLM-as-Critic | arXiv 2025 | - | ||
dVLM-AD |
||||
| dVLM-AD: Enhance Diffusion Vision-Language-Model for Driving via Controllable Reasoning | arXiv 2025 | - | - | |
PLA |
||||
| A Unified Perception-Language-Action Framework for Adaptive Autonomous Driving | arXiv 2025 | - | - | |
AlphaDrive |
||||
| AlphaDrive: Unleashing the Power of VLMs in Autonomous Driving via Reinforcement Learning and Reasoning | arXiv 2025 | - | ||
CoReVLA |
||||
| CoReVLA: A Dual-Stage End-to-End Autonomous Driving Framework for Long-Tail Scenarios via Collect-and-Refine | arXiv 2025 | |||
WAM-Diff |
||||
| WAM-Diff: A Masked Diffusion VLA Framework with MoE and Online Reinforcement Learning for Autonomous Driving | arXiv 2025 | - | ||
VLADriveBench |
||||
| VLADriveBench: Evaluating CoT-Action Relationship in VLA for Autonomous Driving | arXiv 2026 | - | - | |
BLUE |
||||
| BLUE: Toward Better Language Use in Efficient Vision-Language-Action Models for Autonomous Driving | arXiv 2026 | - | - | |
DriveMA |
||||
| DriveMA: Driving Vision-Language-Action Models with verifiable Meta-Actions | CoRL 2026 | |||
C-CoT |
||||
| C-CoT: Counterfactual Chain-of-Thought with Vision-Language Models for Safe Autonomous Driving | arXiv 2026 | - | - | |
MAGNIFIED |
||||
| MAGNIFIED: RL Fine-tuning of Multimodal Large Language Models for Motion Planning | arXiv 2026 | - | - | |
DriveReward |
||||
| DriveReward: A Comprehensive Dataset and Generative Vision-Language Reward Model for Autonomous Driving | arXiv 2026 | - | - | |
nuReasoning |
||||
| nuReasoning: A Reasoning-Centric Dataset and Benchmark for Long-Tail Autonomous Driving | arXiv 2026 | - | - | |
Decision-Making |
||||
| Decision-Making with Lightweight Confidence-Aware Language Model for Autonomous Driving | arXiv 2026 | - | - | |
Is |
||||
| Is VLA Reasoning Faithful? Probing Safety of Chain-of-Causation in Autonomous Driving Models | arXiv 2026 | - | - | |
ReasonBreak |
||||
| ReasonBreak: Probing Vulnerabilities in Reasoning-Enabled Vision-Language-Action Models for Autonomous Driving | arXiv 2026 | - | - | |
Intend, |
||||
| Intend, Reflect, Refine: An Adaptive Multimodal Reflection Framework for Autonomous Driving | arXiv 2026 | - | - | |
Judge, Then Drive |
||||
| Judge, Then Drive: A Critic-Centric Vision Language Action Framework for Autonomous Driving | arXiv 2026 | - | - | |
EvoDrive |
||||
| EvoDrive: Pareto Evolution for Safety-Critical Autonomous Driving via Self-Improving LLM Agents | arXiv 2026 | - | - | |
AutoDrive-P3 |
||||
| AutoDrive-P3: Unified Chain of Perception-Prediction-Planning Thought via Reinforcement Fine-Tuning | ICLR 2026 | - | ||
| Models | ||||
Unifying |
||||
| Unifying Language-Action Understanding and Generation for Autonomous Driving | arXiv 2026 | - | - | |
MindDriver |
||||
| MindDriver: Introducing Progressive Multimodal Reasoning for Autonomous Driving | arXiv 2026 | - | - | |
HERMES |
||||
| HERMES: A Holistic End-to-End Risk-Aware Multimodal Embodied System with Vision-Language Models for Long-Tail Autonomous Driving | arXiv 2026 | - | - | |
Counterfactual VLA |
||||
| Counterfactual VLA: Self-Reflective Vision-Language-Action Model with Adaptive Reasoning | arXiv 2025 | - | - | |
OmniDrive-R1 |
||||
| OmniDrive-R1: Reinforcement-driven Interleaved Multi-modal Chain-of-Thought for Trustworthy Vision-Language Autonomous Driving | arXiv 2025 | - | - | |
BeLLA |
||||
| BeLLA: End-to-End Birds Eye View Large Language Assistant for Autonomous Driving | arXiv 2025 | - | - | |
Reasoning |
||||
| Reasoning-aware Speculative Decoding for Efficient Vision-Language-Action Models in Autonomous Driving | arXiv 2026 | - | - | |
What |
||||
| What Do They See? Interpreting Complex Road Scenarios Through the Eyes of Vision-Language-Action Models for Safe and Trustworthy Autonomous Vehicle Learning | arXiv 2026 | - | - | |
Teaching |
||||
| Teaching Vision-Language-Action Models What to See and Where to Look | arXiv 2026 | - | - | |
Deferred |
||||
| Deferred Exposure of Future Trajectories for Verifiable Reasoning in Autonomous Driving VLMs | arXiv 2026 | - | - | |
Depth |
||||
| Depth-Wise Probing and Pruning of the Planning Token in a Driving Vision-Language-Action Model | arXiv 2026 | - | - | |
CritiqueDriveVLM |
||||
| CritiqueDriveVLM: From Verifier-Guided Reinforcement Learning to Latent Thought Distillation for Autonomous Driving | arXiv 2026 | - | - | |
FactorDrive |
||||
| FactorDrive: Adaptive Multi-Step Reasoning Driven by Planning-Critical Factors for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
XCoT |
||||
| XCoT-VLA: Executable Chain-of-Thought for Vision-Language-Action Driving | arXiv 2026 | - | - | |
MVPruner |
||||
| MVPruner: Dynamic Token Pruning for Accelerating Multi-view Vision-Language Models in Autonomous Driving | arXiv 2026 | - | - | |
FocusDrive |
||||
| FocusDrive: Reasoning with Visual Focus for Autonomous Driving | arXiv 2026 | - | - | |
GeoCoTDrive |
||||
| Explicit Geometric Chain-of-Thought for Vision-Language-Action in Autonomous Driving | arXiv 2026 | - |
:two: Numerical Action Generator
:timer_clock: In chronological order, from the earliest to the latest.
| Model | Paper | Venue | Website | GitHub |
|---|---|---|---|---|
Geo-VLA |
||||
| Geo-VLA: Geometry-Aware Vision-Language-Action Planning via Internalization of Map Semantics | arXiv 2026 | - | - | |
RedLight-VLA |
||||
| RedLight-VLA: Models for traffic-rule grounding and behavioral emphasis in driving policies | arXiv 2026 | - | - | |
Collab-MMI |
||||
| A Collaborative Multi-Modality Interaction for VLA-based End-to-End Autonomous Driving | arXiv 2026 | - | - | |
Drive the Thoughts |
||||
| Drive the Thoughts: Runtime Monitoring of VLA Reasoning-Trajectory Consistency | arXiv 2026 | - | - | |
LMDrive |
||||
| LMDrive: Closed-Loop End-to-End Driving with Large Language Models | CVPR 2024 | |||
BEVDriver |
||||
| BEVDriver: Leveraging BEV Maps in LLMs for Robust Closed-Loop Driving | IROS 2025 | - | - | |
CoVLA-Agent |
||||
| CoVLA: Comprehensive Vision-Language-Action Dataset for Autonomous Driving | WACV 2025 | - | ||
ORION |
||||
| ORION: A Holistic End-to-End Autonomous Driving Framework by Vision-Language Instructed Action Generation | ICCV 2025 | |||
SimLingo |
||||
| SimLingo: Vision-Only Closed-Loop Autonomous Driving with Language-Action Alignment | CVPR 2025 | |||
DriveGPT4-V2 |
||||
| DriveGPT4-V2: Harnessing Large Language Model Capabilities for Enhanced Closed-Loop Autonomous Driving | CVPR 2025 | - | - | |
AutoVLA |
||||
| AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning | NeurIPS 2025 | |||
DriveMoE |
||||
| DriveMoE: Mixture-of-Experts for Vision-Language-Action Model in End-to-End Autonomous Driving | arXiv 2025 | |||
DSDrive |
||||
| DSDrive: Distilling Large Language Model for Lightweight End-to-End Autonomous Driving with Unified Reasoning and Planning | arXiv 2025 | - | - | |
OccVLA |
||||
| OccVLA: Vision-Language-Action Model with Implicit 3D Occupancy Supervision. | arXiv 2025 | - | - | |
VDRive |
||||
| VDRive: Leveraging Reinforced VLA and Diffusion Policy for End-to-End Autonomous Driving | arXiv 2025 | - | - | |
ReflectDrive |
||||
| Discrete Diffusion for Reflective Vision-Language-Action Models in Autonomous Driving | arXiv 2025 | - | ||
E3AD |
||||
| E3AD: An Emotion-Aware Vision-Language-Action Model for Human-Centric End-to-End Autonomous Driving | arXiv 2025 | - | - | |
LCDrive |
||||
| Latent Chain-of-Thought World Modeling for End-to-End Driving | arXiv 2025 | - | - | |
Alpamayo-R1 |
||||
| Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail | arXiv 2025 | - | - | |
UniUGP |
||||
| UniUGP: Unifying understanding, generation, and planing for end-to-end autonomous driving. | arXiv 2025 | - | - | |
MindDrive |
||||
| MindDrive: An All-in-One Framework Bridging World Models and Vision-Language Model for End-to-End Autonomous Driving | arXiv 2025 | - | - | |
AdaThinkDrive |
||||
| AdaThinkDrive: Adaptive Thinking via Reinforcement Learning for Autonomous Driving | arXiv 2025 | - | - | |
Percept-WAM |
||||
| Percept-WAM: Perception-Enhanced World-Awareness-Action Model for Robust End-to-End Autonomous Driving | arXiv 2025 | - | - | |
Reasoning-VLA |
||||
| Reasoning-VLA: A Fast and General Vision-Language-Action Reasoning Model for Autonomous Driving | arXiv 2025 | - | - | |
SpaceDrive |
||||
| SpaceDrive: Infusing Spatial Awareness into VLM-Based Autonomous Driving | arXiv 2025 | - | - | |
OpenDriveVLA |
||||
| OpenDriveVLA: Towards End-to-end Autonomous Driving with Large Vision Language Action Model | AAAI 2026 | |||
WAM-Flow |
||||
| WAM-Flow: Parallel Coarse-to-Fine Motion Planning via Discrete Flow Matching for Autonomous Driving | CVPR 2026 | |||
ColaVLA |
||||
| ColaVLA: Leveraging Cognitive Latent Reasoning for Hierarchical Parallel Trajectory Planning in Autonomous Driving | CVPR 2026 | |||
AutoMoT |
||||
| AutoMoT: A Unified Vision-Language-Action Model with Asynchronous Mixture-of-Transformers for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
OneDrive |
||||
| OneDrive: Unified Multi-Paradigm Driving with Vision-Language-Action Models | arXiv 2026 | - | - | |
UniDriveVLA |
||||
| UniDriveVLA: Unifying Understanding, Perception, and Action Planning for Autonomous Driving | arXiv 2026 | - | - | |
VLA-World |
||||
| Learning Vision-Language-Action World Models for Autonomous Driving | arXiv 2026 | - | - | |
Metis |
||||
| Metis: A Generalizable and Efficient World-Action Model for Autonomous Driving and Urban Navigation | arXiv 2026 | - | - | |
MindVLA-U1 |
||||
| MindVLA-U1: VLA Beats VA with Unified Streaming Architecture for Autonomous Driving | arXiv 2026 | - | - | |
DVGT-2 |
||||
| DVGT-2: Vision-Geometry-Action Model for Autonomous Driving at Scale | arXiv 2026 | - | - | |
VLGA |
||||
| VLGA: Vision-Language-Geometry-Action Models for Autonomous Driving | arXiv 2026 | - | - | |
VECTOR-Drive |
||||
| VECTOR-Drive: Tightly Coupled Vision-Language and Trajectory Expert Routing for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
ChainFlow-VLA |
||||
| ChainFlow-VLA: Causal Flow Planning with Vision-Language Models | arXiv 2026 | - | - | |
LVDrive |
||||
| LVDrive: Latent Visual Representation Enhanced Vision-Language-Action Autonomous Driving Model | arXiv 2026 | - | - | |
StyleVLA |
||||
| StyleVLA: Driving Style-Aware Vision Language Action Model for Autonomous Driving | arXiv 2026 | - | - | |
Masked-VLA-Diffusion |
||||
| Efficient and Explainable End-to-End Autonomous Driving via Masked Vision-Language-Action Diffusion | arXiv 2026 | - | - | |
Uni-World VLA |
||||
| Uni-World VLA: Interleaved World Modeling and Planning for Autonomous Driving | arXiv 2026 | - | - | |
Does |
||||
| Does Visual Information Play a Decisive Role in Vision-Language-Action Model Driving Behavior? | arXiv 2026 | - | - | |
Reasoning |
||||
| Reasoning About Traversability: Language-Guided Off-Road 3D Trajectory Planning | arXiv 2026 | - | - | |
SpanVLA |
||||
| SpanVLA: Efficient Action Bridging and Learning from Negative-Recovery Samples for Vision-Language-Action Model | arXiv 2026 | - | - | |
Sim2Real-AD |
||||
| Sim2Real-AD: A Modular Sim-to-Real Framework for Deploying VLM-Guided Reinforcement Learning in Real-World Autonomous Driving | arXiv 2026 | - | - | |
Drive My Way |
||||
| Drive My Way: Preference Alignment of Vision-Language-Action Model for Personalized Driving | arXiv 2026 | - | - | |
DriveVLM-RL |
||||
| DriveVLM-RL: Neuroscience-Inspired Reinforcement Learning with Vision-Language Models for Safe and Deployable Autonomous Driving | arXiv 2026 | - | - | |
Learning from Mistakes |
||||
| Learning from Mistakes: Post-Training for Driving VLA with Takeover Data | arXiv 2026 | - | - | |
SAMoE-VLA |
||||
| SAMoE-VLA: A Scene Adaptive Mixture-of-Experts Vision-Language-Action Model for Autonomous Driving | arXiv 2026 | - | - | |
LaST-VLA |
||||
| LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving | arXiv 2026 | - | - | |
PixelPilot |
||||
| PixelPilot: Scalable Vision-Language-Action Models for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
S-squared |
||||
| S-squared-VLA: Decoupling Semantic and Spatial Streams in Vision-Language-Action Models for Autonomous Driving | arXiv 2026 | - | - | |
WAM |
||||
| WAM-Diff2: Hierarchical AR-to-Diffusion Distillation for Highly Efficient Autonomous Driving VLA | arXiv 2026 | - | - | |
AnchorVLA |
||||
| AnchorVLA: Bridging Discrete Decisions and Continuous Trajectories for Vision-Language-Action Planning | arXiv 2026 | - | - | |
CLEAR |
||||
| CLEAR: Closed-Loop Reinforcement Learning at Scale for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
Post |
||||
| Post-Training in End-to-End Autonomous Driving | arXiv 2026 | - | - | |
Latent |
||||
| Latent-Centroid Steering: Single-Pass Classifier-Free Guidance for Command-Aligned Autonomous Driving | arXiv 2026 | - | - | |
Inference |
||||
| Inference-Time Attention Steering for Vision-Language-Action Driving Models | arXiv 2026 | - | - | |
Geo-VLA |
||||
| Geo-VLA: Geometry-Aware Vision-Language-Action Planning via Internalization of Map Semantics | arXiv 2026 | - | - | |
RedLight-VLA |
||||
| RedLight-VLA: Models for traffic-rule grounding and behavioral emphasis in driving policies | arXiv 2026 | - | - | |
Collaborative Multi-Modality |
||||
| A Collaborative Multi-Modality Interaction for VLA-based End-to-End Autonomous Driving | arXiv 2026 | - | - | |
Qwen-Drive-1.0 |
||||
| Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving | arXiv 2026 | |||
LaPla |
||||
| Continuous Actions from Discrete Minds: Latent-Aligned Planning for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
GRAVA |
||||
| GRAVA: Grounded Reasoning-to-Action Representation and Learning for Autonomous Driving | arXiv 2026 | - | ||
RAF-VLA |
||||
| RAF-VLA: Representation Alignment with the Future for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
FIVE-VLA |
||||
| FIVE-VLA: Fast and EffectIVE Autonomous Driving with Recurrent Action Memory | arXiv 2026 | - | - | |
Run-then-Walk |
||||
| Sometimes You Gotta Run Before You Can Walk: Run-then-Walk Scheduling Strategy for VLM Autonomous Driving | arXiv 2026 | - | ||
| Models | ||||
CAR-VLA |
||||
| CAR-VLA: Complexity-Aware and Risk-Adaptive Reasoning for Autonomous Driving | arXiv 2026 | - | ||
RefineDrive |
||||
| RefineDrive: Reliable Failure-Guided Learning for Vision-Language-Action Driving | arXiv 2026 | - | - | |
AD-Memo |
||||
| Vision-Language-Action Autonomous Driving Agent with Language-based Memory | arXiv 2026 | - | - | |
AUGUstus |
||||
| AUGUstus: Autonomous Driving VLA Model with Unified Generation and Understanding | arXiv 2026 | |||
| Models | ||||
YOUDrive |
||||
| YOUDrive: Driving Style as a Composable Direction for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
TacitVLA |
||||
| TacitVLA: Reasoning Vision-Language-Action Models Without Reasoning Latency | arXiv 2026 | - | ||
DriveWLA |
||||
| DriveWLA: Unifying Future Dynamics and Scene Semantics through World-Language-Action Modeling for Autonomous Driving | arXiv 2026 | - | - | |
Reflect-VLA |
||||
| Reflect-VLA: Internalizing Reflective Guidance via Dual-Path On-Policy Self-Distillation for Autonomous Driving | arXiv 2026 | - | ||
AffordDrive3D |
||||
| AffordDrive3D: Affordance-Aware World-Action Modeling with Spatial Understanding | arXiv 2026 | - | ||
:three: Explicit Action Guidance
:timer_clock: In chronological order, from the earliest to the latest.
| Model | Paper | Venue | Website | GitHub |
|---|---|---|---|---|
DriveVLM |
||||
| DriveVLM: The Convergence of Autonomous Driving and Large Vision-Language Models | CoRL 2024 | - | ||
LeapAD |
||||
| Continuously Learning, Adapting, and Improving: A Dual-Process Approach to Autonomous Driving | NeurIPS 2024 | |||
FasionAD |
||||
| FASIONAD: Fast and Slow Fusion Thinking Systems for Human-Like Autonomous Driving with Adaptive Feedback | arXiv 2024 | - | - | |
Senna |
||||
| Senna: Bridging Large Vision-Language Models and End-to-End Autonomous Driving | arXiv 2024 | - | ||
DualAD |
||||
| DualAD: Dual-Layer Planning for Reasoning in Autonomous Driving | IROS 2025 | |||
DME-Driver |
||||
| DME-Driver: Integrating Human Decision Logic and 3D Scene Perception in Autonomous Driving | AAAI 2025 | - | - | |
SOLVE |
||||
| SOLVE: Synergy of Language-Vision and End-to-End Networks for Autonomous Driving | CVPR 2025 | - | - | |
ReAL-AD |
||||
| ReAL-AD: Towards Human-Like Reasoning in End-to-End Autonomous Driving | ICCV 2025 | - | ||
LeapVAD |
||||
| LeapVAD: A Leap in Autonomous Driving via Cognitive Perception and Dual-Process Thinking | TNNLS 2025 | - | - | |
DiffVLA |
||||
| DiffVLA: Vision-Language Guided Diffusion Planning for Autonomous Driving | arXiv 2025 | - | - | |
FasionAD++ |
||||
| FASIONAD++: Integrating High-Level Instruction and Information Bottleneck in Fast-Slow fusion Systems for Enhanced Safety in Autonomous Driving with Adaptive Feedback | arXiv 2025 | - | - | |
HiST-VLA |
||||
| HiST-VLA: A Hierarchical Spatio-Temporal Vision-Language-Action Model for End-to-End Autonomous Driving | arXiv 2026 | - | - | |
Senna-2 |
||||
| Senna-2: Aligning VLM and End-to-End Driving Policy for Consistent Decision Making and Planning | arXiv 2026 | - | - | |
From |
||||
| From Representational Complementarity to Dual Systems: Synergizing VLM and Vision-Only Backbones for End-to-End Driving | arXiv 2026 | - | - | |
Fast-dDrive |
||||
| Fast-dDrive: Efficient Block-Diffusion VLM for Autonomous Driving | arXiv 2026 | - | - | |
DiffVLA++ |
||||
| DiffVLA++: Bridging Cognitive Reasoning and End-to-End Driving through Metric-Guided Alignment | arXiv 2025 | - | - | |
RT-VLA |
||||
| RT-VLA: Real-Time Vision-Language-Action Models via Knowledge Distillation | arXiv 2026 | - | - | |
Slow |
||||
| Slow Brain, Fast Planner: Latency-Resilient VLM-Augmented Urban Navigation | arXiv 2026 | - | - | |
SimpleVSF |
||||
| SimpleVSF: VLM-Scoring Fusion for Trajectory Prediction of End-to-End Autonomous Driving | arXiv 2025 | - | - | |
CoWorld-VLA |
||||
| CoWorld-VLA: Thinking in a Multi-Expert World Model for Autonomous Driving | arXiv 2026 | - | - | |
OneVL |
||||
| Xiaomi OneVL: One-Step Latent Reasoning and Planning with Vision-Language Explanation | arXiv 2026 | - | - | |
WorldVLM |
||||
| WorldVLM: Combining World Model Forecasting and Vision-Language Reasoning | arXiv 2026 | - | - | |
NaviDriveVLM |
||||
| NaviDriveVLM: Decoupling High-Level Reasoning and Motion Planning for Autonomous Driving | arXiv 2026 | - | - | |
PRAM-R |
||||
| PRAM-R: A Perception-Reasoning-Action-Memory Framework with LLM-Guided Modality Routing for Adaptive Autonomous Driving | arXiv 2026 | - | - | |
CorrectAD |
||||
| CorrectAD: A Self-Correcting Agentic System to Improve End-to-end Planning in Autonomous Driving | arXiv 2025 | - | - | |
MTRDrive |
||||
| MTRDrive: Memory-Tool Synergistic Reasoning for Robust Autonomous Driving in Corner Cases | arXiv 2025 | - | - | |
Think |
||||
| Think at 5 Hz, Act at 20 Hz: Asynchronous Fast-Slow Vision-Language-Action Inference for Closed-Loop Driving | arXiv 2026 | - | - | |
DriveVLA |
||||
| DriveVLA-M0: Failure-Aware Memory Augmentation for Autonomous Driving | arXiv 2026 | - | - | |
FIRE |
||||
| FIRE-VLA: Failure-Informed Self-Evolution for Vision-Language-Action Models in Autonomous Driving | arXiv 2026 | - | - | |
FlashDrive |
||||
| FlashDrive: Flash Vision-Language-Action Inference for Autonomous Driving | arXiv 2026 | - | - | |
GAPL |
||||
| GAPL: Grounded Action-effect Policy Learning for LLM-Based Trajectory Planning | arXiv 2026 | - | - | |
Multi-Agent LLM Orchestration |
||||
| Multi-Agent Orchestration with the Common-Sense Reasoning Capabilities of LLMs for Autonomous Driving | arXiv 2026 | - | - | |
LADA |
||||
| Less Language, More Latents: Annotation-Efficient VLAs for Driving | arXiv 2026 | - | - | |
RCVLA |
||||
| RCVLA: 4D Radar-Grounded Semantic Reasoning and Trajectory Arbitration for Autonomous Driving | arXiv 2026 | - | - | |
AutoAgent0 |
||||
| AutoAgent0: An Agentic Runtime for Safe Closed-Loop Driving | arXiv 2026 | |||
SIGMA |
||||
| Not All Uncertainty Matters: Simulation-in-the-Loop Fast-Slow Reasoning for Decision-Critical Autonomous Driving System | arXiv 2026 | - | - | |
:four: Implicit Representations Transfer
:timer_clock: In chronological order, from the earliest to the latest.
| Model | Paper | Venue | Website | GitHub |
|---|---|---|---|---|
VLP |
||||
| VLP: Vision Language Planning for Autonomous Driving | CVPR 2024 | - | - | |
VLM-AD |
||||
| VLM-AD: End-to-End Autonomous Driving through Vision-Language Model Supervision | CoRL 2025 | - | - | |
DiMA |
||||
| Distilling Multi-modal Large Language Models for Autonomous Driving | CVPR 2025 | - | - | |
DINO-Foresight |
||||
| DINO-Foresight: Looking into the Future with DINO | NeurIPS 2025 | |||
ALN-P3 |
||||
| ALN-P3: Unified Language Alignment for Perception, Prediction, and Planning in Autonomous Driving | arXiv 2025 | - | - | |
VERDI |
||||
| VERDI: VLM-Embedded Reasoning for Autonomous Driving | arXiv 2025 | - | - | |
VLM-E2E |
||||
| VLM-E2E: Enhancing End-to-End Autonomous Driving with Multimodal Driver Attention Fusion | arXiv 2025 | - | - | |
ReCogDrive |
||||
| ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Driving | arXiv 2025 | |||
InsightDrive |
||||
| InsightDrive: Insight Scene Representation for End-to-End Autonomous Driving | arXiv 2025 | - | ||
NetRoller |
||||
| NetRoller: Interfacing General and Specialized Models for End-to-End Autonomous Driving | arXiv 2025 | - | ||
ViLaD |
||||
| ViLaD: A Large Vision Language Diffusion Framework for End-to-End Autonomous Driving | arXiv 2025 | - | - | |
OmniScene |
||||
| OmniScene: Attention-Augmented Multimodal 4D Scene Understanding for Autonomous Driving | arXiv 2025 | - | - | |
LMAD |
||||
| LMAD: Integrated End-to-End VisionLanguage Model for Explainable Autonomous Driving | arXiv 2025 | - | - | |
BEVLM |
||||
| BEVLM: Distilling Semantic Knowledge from LLMs into Bird's-Eye View Representations | ECCV 2026 | - | ||
Outcome |
||||
| Outcome-Guided Distillation: A Teacher-Student Framework to Advance VLM Reasoning in Autonomous Driving | arXiv 2026 | - | - | |
Relational Latent World Models |
||||
| Relationally Grounded Latent World Models for Autonomous Driving | arXiv 2026 | - | - | |
3. Datasets & Benchmarks
:timer_clock: In chronological order, from the earliest to the latest.
:one: Vision-Action Datasets
| Dataset | Paper | Venue | Website | GitHub |
|---|---|---|---|---|
SkyDrive |
||||
| SkyDrive: Learning to Drive in a New City from Aerial Traffic Monitoring | arXiv 2026 | - | - | |
BDD100K |
||||
| BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning | CVPR 2020 | |||
nuScenes |
||||
| nuScenes: A Multimodal Dataset for Autonomous Driving | CVPR 2020 | - | ||
Waymo |
||||
| Scalability in Perception for Autonomous Driving: Waymo Open Dataset | CVPR 2020 | |||
nuPlan |
||||
| nuPlan: A Closed-Loop ML-Based Planning Benchmark for Autonomous Vehicles | arXiv 2021 | |||
Argoverse 2 |
||||
| Argoverse 2: Next Generation Datasets for Self-Driving Perception and Forecasting | NeurIPS 2021 | |||
Bench2Drive |
||||
| Bench2Drive: Towards Multi-Ability Benchmarking of Closed-Loop End-to-End Autonomous Driving | NeurIPS 2024 | - | ||
RoboBEV |
||||
| Benchmarking and Improving Bird's Eye View Perception Robustness in Autonomous Driving | TPAMI 2025 | - | ||
WOD-E2E |
||||
| WOD-E2E: Waymo Open Dataset for End-to-End Driving in Challenging Long-Tail Scenarios | arXiv 2025 | |||
navdream |
||||
| The Constant Eye: Benchmarking and Bridging Appearance Robustness in Autonomous Driving | arXiv 2026 | - | - | |
SIREN-Bench |
||||
| SIREN-Bench: Behavior-Driven Generation and Evaluation of Emergency-Vehicle Interactions | arXiv 2026 | - | - | |
SkyDrive |
[![arX |