mlsanityData to 7 Oct 2026

AlphaDrive: Unleashing the Power of VLMs in Autonomous Driving via Reinforcement Learning and Reasoning

arXiv 2503.07608 ↗First posted 10 Mar 2025cs.CV cs.ROcited by 108

On the map: LLM reasoning → Chain-of-thought reasoning for driving

also in 1 more
Abstract (arXiv)

OpenAI o1 and DeepSeek R1 achieve or even surpass human expert-level performance in complex domains like mathematics and science, with reinforcement learning (RL) and reasoning playing a crucial role. In autonomous driving, recent end-to-end models have greatly improved planning performance but still struggle with long-tailed problems due to limited common sense and reasoning abilities. Some studies integrate vision-language models (VLMs) into autonomous driving, but they typically rely on pre-trained models with simple supervised fine-tuning (SFT) on driving data, without further exploration of training strategies or optimizations specifically tailored for planning. In this paper, we propose AlphaDrive, a RL and reasoning framework for VLMs in autonomous driving. AlphaDrive introduces four GRPO-based RL rewards tailored for planning and employs a two-stage planning reasoning training strategy that combines SFT with RL. As a result, AlphaDrive significantly improves both planning performance and training efficiency compared to using only SFT or without reasoning. Moreover, we are also excited to discover that, following RL training, AlphaDrive exhibits some emergent multimodal planning capabilities, which is critical for improving driving safety and efficiency. To the best of our knowledge, AlphaDrive is the first to integrate GRPO-based RL with planning reasoning into autonomous driving. Code will be released to facilitate future research.

Key facts. Cited by 108 papers in the map (data through 7 Oct 2026). Placed in Chain-of-thought reasoning for driving, one of 17 directions in LLM reasoning. Builds on 43 of its 49 references that are in the map. First posted on arXiv 10 Mar 2025.

Builds on

43 of 49 references.

References · 43

  1. 2017Attention Is All You Needcited by 59,932
  2. 2020Language Models are Few-Shot Learnerscited by 32,199
  3. 2023GPT-4 Technical Reportcited by 18,132
  4. 2022Training language models to follow instructions with human feedbackcited by 15,712
  5. 2017Proximal Policy Optimization Algorithmscited by 14,441
  6. 2024The Llama 3 Herd of Modelscited by 14,192
  7. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 13,958
  8. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  9. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  10. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  11. 2023Visual Instruction Tuningcited by 7,792
  12. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
31 more
  1. 2023BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Modelscited by 6,069
  2. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  3. 2023Improved Baselines with Visual Instruction Tuningcited by 4,362
  4. 2022Flamingo: a Visual Language Model for Few-Shot Learningcited by 4,337
  5. 2022BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generationcited by 4,310
  6. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,151
  7. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  8. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  9. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Taskscited by 2,534
  10. 2024Qwen2 Technical Reportcited by 2,076
  11. 2014CIDEr: Consensus-based Image Description Evaluationcited by 2,027
  12. 2024Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scalingcited by 1,573
  13. 2022Planning-oriented Autonomous Drivingcited by 1,068
  14. 2023VAD: Vectorized Scene Representation for Efficient Autonomous Drivingcited by 613
  15. 2023DriveLM: Driving with Graph Visual Question Answeringcited by 527
  16. 2024DriveVLM: The Convergence of Autonomous Driving and Large Vision-Language Modelscited by 482
  17. 2023DriveGPT4: Interpretable End-to-end Autonomous Driving via Large Language Modelcited by 455
  18. 2021Multi-Modal Fusion Transformer for End-to-End Autonomous Drivingcited by 372
  19. 2024DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Drivingcited by 324
  20. 2024VADv2: End-to-End Vectorized Autonomous Driving via Probabilistic Planningcited by 287
  21. 2023Mathematical Capabilities of ChatGPTcited by 263
  22. 2023NuScenes-QA: A Multi-modal Visual Question Answering Benchmark for Autonomous Driving Scenariocited by 258
  23. 2016A Survey of Motion Planning and Control Techniques for Self-driving Urban Vehiclescited by 251
  24. 2023Driving with LLMs: Fusing Object-Level Vector Modality for Explainable Autonomous Drivingcited by 231
  25. 2023Self-Evaluation Guided Beam Search for Reasoningcited by 195
  26. 2021Monte Carlo Tree Search: A Review of Recent Modifications and Applicationscited by 158
  27. 2024Senna: Bridging Large Vision-Language Models and End-to-End Autonomous Drivingcited by 156
  28. 2023Language Prompt for Autonomous Drivingcited by 110
  29. 2025RAD: Training an End-to-End Driving Policy via Large-Scale 3DGS-based Reinforcement Learningcited by 76
  30. 2024Embodied Understanding of Driving Scenarioscited by 63
  31. 2024Diffusion-VLA: Generalizable and Interpretable Robot Foundation Model via Self-Generated Reasoningcited by 57

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 40; 2026: 68.

108 papers.

Sort:

By area:

  1. 2025AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuningcited by 252
  2. 2025ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Drivingcited by 190
  3. 2025End-to-End Driving with Online Trajectory Evaluation via BEV World Modelcited by 145
  4. 2025A Survey on Vision-Language-Action Models: An Action Tokenization Perspectivecited by 96
  5. 2025AdaThinkDrive: Adaptive Thinking via Reinforcement Learning for Autonomous Drivingcited by 57
  6. 2025DiffusionDriveV2: Reinforcement Learning-Constrained Truncated Diffusion Modeling in End-to-End Autonomous Drivingcited by 52
  7. 2025DriveDPO: Policy Learning via Safety DPO For End-to-End Autonomous Drivingcited by 39
  8. 2025Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learningcited by 38
  9. 2025AgentThink: A Unified Framework for Tool-Augmented Chain-of-Thought Reasoning in Vision-Language Models for Autonomous Drivingcited by 35
  10. 2025AutoDrive-R2: Incentivizing Reasoning and Self-Reflection Capacity for VLA Model in Autonomous Drivingcited by 32
  11. 2025Vision-Language-Action Models for Autonomous Driving: Past, Present, and Futurecited by 32
  12. 2025ChestX-Reasoner: Advancing Radiology Foundation Models with Reasoning through Step-by-Step Verificationcited by 30
38 more
  1. 2025ReconDreamer-RL: Enhancing Reinforcement Learning via Diffusion-based Scene Reconstructioncited by 27
  2. 2026LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Drivingcited by 26
  3. 2025LangCoop: Collaborative Driving with Languagecited by 25
  4. 2025MindDrive: A Vision-Language-Action Model for Autonomous Driving via Online Reinforcement Learningcited by 25
  5. 2025Video-STAR: Reinforcing Open-Vocabulary Action Recognition with Toolscited by 23
  6. 2025Counterfactual VLA: Self-Reflective Vision-Language-Action Model with Adaptive Reasoningcited by 21
  7. 2025A Survey on Agentic Multimodal Large Language Modelscited by 19
  8. 2025WorldRFT: Latent World Model Planning with Reinforcement Fine-Tuning for Autonomous Drivingcited by 19
  9. 2026VGGDrive: Empowering Vision-Language Models with Cross-View Geometric Grounding for Autonomous Drivingcited by 19
  10. 2026DriveFine: Refining-Augmented Masked Diffusion VLA for Precise and Robust Drivingcited by 18
  11. 2025DSDrive: Distilling Large Language Model for Lightweight End-to-End Autonomous Driving with Unified Reasoning and Planningcited by 17
  12. 2025DIVER: Reinforced Diffusion Breaks Imitation Bottlenecks in End-to-End Autonomous Drivingcited by 17
  13. 2026Unleashing VLA Potentials in Autonomous Driving via Explicit Learning from Failurescited by 17
  14. 2025WAM-Flow: Parallel Coarse-to-Fine Motion Planning via Discrete Flow Matching for Autonomous Drivingcited by 15
  15. 2026SpanVLA: Efficient Action Bridging and Learning from Negative-Recovery Samples for Vision-Language-Action Modelcited by 15
  16. 2025HMVLM: Multistage Reasoning-Enhanced Vision-Language Model for Long-Tailed Driving Scenarioscited by 14
  17. 2026On-the-Fly VLA Adaptation via Test-Time Reinforcement Learningcited by 14
  18. 2026AutoDrive-P3: Unified Chain of Perception-Prediction-Planning Thought via Reinforcement Fine-Tuningcited by 13
  19. 2025Chain-of-Thought for Autonomous Driving: A Comprehensive Survey and Future Prospectscited by 12
  20. 2025EvaDrive: Evolutionary Adversarial Policy Optimization for End-to-End Autonomous Drivingcited by 12
  21. 2026DVGT-2: Vision-Geometry-Action Model for Autonomous Driving at Scalecited by 12
  22. 2025Poutine: Vision-Language-Trajectory Pre-Training and Reinforcement Learning Post-Training Enable Robust End-to-End Autonomous Drivingcited by 11
  23. 2025DriveAgent-R1: Advancing VLM-based Autonomous Driving with Active Perception and Hybrid Thinkingcited by 11
  24. 2025OpenREAD: Reinforced Open-Ended Reasoning for End-to-End Autonomous Driving with LLM-as-Criticcited by 11
  25. 2026Learning Vision-Language-Action World Models for Autonomous Drivingcited by 11
  26. 2025UniUGP: Unifying Understanding, Generation, and Planing For End-to-end Autonomous Drivingcited by 10
  27. 2026FLARE: Learning Future-Aware Latent Representations from Vision-Language Models for Autonomous Drivingcited by 9
  28. 2026Unifying Language-Action Understanding and Generation for Autonomous Drivingcited by 9
  29. 2026AutoMoT: A Unified Vision-Language-Action Model with Asynchronous Mixture-of-Transformers for End-to-End Autonomous Drivingcited by 9
  30. 2026Learning from Mistakes: Post-Training for Driving VLA with Takeover Datacited by 8
  31. 2026Bench2Drive-VL: Benchmarks for Closed-Loop Autonomous Driving with Vision-Language Modelscited by 8
  32. 2026RAD-2: Scaling Reinforcement Learning in a Generator-Discriminator Frameworkcited by 7
  33. 2026Generative Scenario Rollouts for End-to-End Autonomous Drivingcited by 6
  34. 2026Vega: Learning to Drive with Natural Language Instructionscited by 6
  35. 2025DriveRX: A Vision-Language Reasoning Model for Cross-Task Autonomous Drivingcited by 5
  36. 2025CoC-VLA: Delving into Adversarial Domain Transfer for Explainable Autonomous Driving via Chain-of-Causality Visual-Language-Action Modelcited by 5
  37. 2025Spatial Retrieval Augmented Autonomous Drivingcited by 5
  38. 2026HiST-VLA: A Hierarchical Spatio-Temporal Vision-Language-Action Model for End-to-End Autonomous Drivingcited by 5

50 most cited of 108 All 108 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025COVLM-RL: Critical Object-Oriented Reasoning for Autonomous Driving Using VLM-Guided Reinforcement Learning
  2. 2026NaviDriveVLM: Decoupling High-Level Reasoning and Motion Planning for Autonomous Driving
  3. 2026Found-RL: foundation model-enhanced reinforcement learning for autonomous driving
  4. 2024VLM-RL: A Unified Vision Language Models and Reinforcement Learning Framework for Safe Autonomous Driving
  5. 2026ThinkDrive: Chain-of-Thought Guided Progressive Reinforcement Learning Fine-Tuning for Autonomous Driving
  6. 2025VERDI: VLM-Embedded Reasoning for Autonomous Driving
  7. 2025VLAD: A VLM-Augmented Autonomous Driving Framework with Hierarchical Planning and Interpretable Decision Process
  8. 2026MindDriver: Introducing Progressive Multimodal Reasoning for Autonomous Driving