mlsanityData to 7 Oct 2026

DiffusionDriveV2: Reinforcement Learning-Constrained Truncated Diffusion Modeling in End-to-End Autonomous Driving

arXiv 2512.07745 ↗First posted 8 Dec 2025cs.CVcited by 52

On the map: Diffusion models → Diffusion and flow policies for robotics

also in 1 more
Abstract (arXiv)

Generative diffusion models for end-to-end autonomous driving often suffer from mode collapse, tending to generate conservative and homogeneous behaviors. While DiffusionDrive employs predefined anchors representing different driving intentions to partition the action space and generate diverse trajectories, its reliance on imitation learning lacks sufficient constraints, resulting in a dilemma between diversity and consistent high quality. In this work, we propose DiffusionDriveV2, which leverages reinforcement learning to both constrain low-quality modes and explore for superior trajectories. This significantly enhances the overall output quality while preserving the inherent multimodality of its core Gaussian Mixture Model. First, we use scale-adaptive multiplicative noise, ideal for trajectory planning, to promote broad exploration. Second, we employ intra-anchor GRPO to manage advantage estimation among samples generated from a single anchor, and inter-anchor truncated GRPO to incorporate a global perspective across different anchors, preventing improper advantage comparisons between distinct intentions (e.g., turning vs. going straight), which can lead to further mode collapse. DiffusionDriveV2 achieves 91.2 PDMS on the NAVSIM v1 dataset and 85.5 EPDMS on the NAVSIM v2 dataset in closed-loop evaluation with an aligned ResNet-34 backbone, setting a new record. Further experiments validate that our approach resolves the dilemma between diversity and consistent high quality for truncated diffusion models, achieving the best trade-off. Code and model will be available at https://github.com/hustvl/DiffusionDriveV2

Key facts. Cited by 52 papers in the map (data through 7 Oct 2026). Placed in Diffusion and flow policies for robotics, one of 14 directions in Diffusion models. Builds on 49 of its 57 references that are in the map. First posted on arXiv 8 Dec 2025.

Builds on

49 of 57 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025DIVER: Reinforced Diffusion Breaks Imitation Bottlenecks in End-to-End Autonomous Driving5 Jul
  2. 2025GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation in End-to-End Autonomous Driving7 Mar
  3. 2024DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving22 Nov
  4. 2024Diffusion Policy Policy Optimization1 Sep
  5. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusion7 Mar

Other references · 44

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  3. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  4. 2017Proximal Policy Optimization Algorithmscited by 14,441
  5. 2020Denoising Diffusion Implicit Modelscited by 9,655
  6. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  7. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,185
  8. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  9. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,141
  10. 2017CARLA: An Open Urban Driving Simulatorcited by 2,431
  11. 2022Planning-oriented Autonomous Drivingcited by 1,068
  12. 2022Planning with Diffusion for Flexible Behavior Synthesiscited by 1,023
32 more
  1. 2021ByteTrack: Multi-Object Tracking by Associating Every Detection Boxcited by 812
  2. 2021BEVDet: High-performance Multi-camera 3D Object Detection in Bird-Eye-Viewcited by 711
  3. 2023VAD: Vectorized Scene Representation for Efficient Autonomous Drivingcited by 613
  4. 2020FairMOT: On the Fairness of Detection and Re-Identification in Multiple Object Trackingcited by 489
  5. 2021NuPlan: A closed-loop ML-based planning benchmark for autonomous vehiclescited by 485
  6. 2022TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Drivingcited by 477
  7. 2019MultiPath: Multiple Probabilistic Anchor Trajectory Hypotheses for Behavior Predictioncited by 417
  8. 2024NAVSIM: Data-Driven Non-Reactive Autonomous Vehicle Simulation and Benchmarkingcited by 343
  9. 2022MapTR: Structured Modeling and Learning for Online Vectorized HD Map Constructioncited by 311
  10. 2022BEVFormer v2: Adapting Modern Image Backbones to Bird's-Eye-View Recognition via Perspective Supervisioncited by 302
  11. 2024VADv2: End-to-End Vectorized Autonomous Driving via Probabilistic Planningcited by 287
  12. 2021MultiPath++: Efficient Information Fusion and Trajectory Aggregation for Behavior Predictioncited by 246
  13. 2024Hydra-MDP: End-to-end Multimodal Planning with Multi-target Hydra-Distillationcited by 244
  14. 2019CenterMask : Real-Time Anchor-Free Instance Segmentationcited by 200
  15. 2023MapTRv2: An End-to-End Framework for Online Vectorized HD Map Constructioncited by 192
  16. 2025ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Drivingcited by 190
  17. 2025End-to-End Driving with Online Trajectory Evaluation via BEV World Modelcited by 145
  18. 2025Pseudo-Simulation for Autonomous Drivingcited by 141
  19. 2022Imitation Is Not Enough: Robustifying Imitation with Reinforcement Learning for Challenging Driving Scenarioscited by 124
  20. 2021Learning to drive from a world on railscited by 114
  21. 2025AlphaDrive: Unleashing the Power of VLMs in Autonomous Driving via Reinforcement Learning and Reasoningcited by 108
  22. 2025Hydra-MDP++: Advancing End-to-End Driving via Expert-Guided Hydra-Distillationcited by 93
  23. 2025DriveSuprim: Towards Precise Trajectory Selection for End-to-End Planningcited by 91
  24. 2024DRAMA: An Efficient End-to-end Motion Planner for Autonomous Driving with Mambacited by 87
  25. 2025RAD: Training an End-to-End Driving Policy via Large-Scale 3DGS-based Reinforcement Learningcited by 76
  26. 2023UniPAD: A Universal Pre-training Paradigm for Autonomous Drivingcited by 73
  27. 2025ARTEMIS: Autoregressive End-to-End Trajectory Planning with Mixture of Experts for Autonomous Drivingcited by 66
  28. 2025Generalized Trajectory Scoring for End-to-end Multimodal Planningcited by 63
  29. 2024Solving Motion Planning Tasks with a Scalable Generative Modelcited by 40
  30. 2024VisionPAD: A Vision-Centric Pre-training Paradigm for Autonomous Drivingcited by 15
  31. 2025ResAD: Normalized Residual Trajectory Modeling for End-to-End Autonomous Drivingcited by 12
  32. 2024MIM4D: Masked Modeling with Multi-View Video for Autonomous Driving Representation Learningcited by 8

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 1; 2026: 51.

52 papers.

Sort:

By area:

  1. 2025Vision-Language-Action Models for Autonomous Driving: Past, Present, and Futurecited by 32
  2. 2026SparseDriveV2: Scoring is All You Need for End-to-End Autonomous Drivingcited by 21
  3. 2026DriveFine: Refining-Augmented Masked Diffusion VLA for Precise and Robust Drivingcited by 18
  4. 2026SpanVLA: Efficient Action Bridging and Learning from Negative-Recovery Samples for Vision-Language-Action Modelcited by 15
  5. 2026DriveFuture: Future-Aware Latent World Models for Autonomous Drivingcited by 15
  6. 2026EponaV2: Driving World Model with Comprehensive Future Reasoningcited by 12
  7. 2026ExploreVLA: Dense World Modeling and Exploration for End-to-End Autonomous Drivingcited by 10
  8. 2026BridgeSim: Unveiling the OL-CL Gap in End-to-End Autonomous Drivingcited by 9
  9. 2026CLOVER: Closed-Loop Value Estimation and Ranking for End-to-End Autonomous Driving Planningcited by 9
  10. 2026RAD-2: Scaling Reinforcement Learning in a Generator-Discriminator Frameworkcited by 7
  11. 2026World Engine: Towards the Era of Post-Training for Autonomous Drivingcited by 6
  12. 2026Senna-2: Aligning VLM and End-to-End Driving Policy for Consistent Decision Making and Planningcited by 4
38 more
  1. 2026HAD: Combining Hierarchical Diffusion with Metric-Decoupled RL for End-to-End Drivingcited by 4
  2. 2026MDrive: Benchmarking Closed-Loop Cooperative Driving for End-to-End Multi-agent Systemscited by 4
  3. 2026EvoDriveVLA: Evolving Driving VLA Models via Collaborative Perception-Planning Distillationcited by 3
  4. 2026Discrete-WAM: Unified Discrete Vision-Action Token Editing for World-Policy Learningcited by 3
  5. 2026CRAFT: Counterfactual-to-Interactive Reinforcement Fine-Tuning for Driving Policiescited by 2
  6. 2026DriveMA: Driving Vision-Language-Action Models with verifiable Meta-Actionscited by 2
  7. 2026IDOL: Inverse-Dynamics-Guided Future Prediction for End-to-End Autonomous Drivingcited by 2
  8. 2026CLEAR: Cognition and Latent Evaluation for Adaptive Routing in End-to-End Autonomous Drivingcited by 2
  9. 2026DriveJudge: Rethinking Autonomous Driving Evaluation with Vision-Language Modelscited by 2
  10. 2026ForgeDrive: Bidirectional Cross-Conditioning for Unified Visual-Action Generation in Autonomous Drivingcited by 2
  11. 2026WA-JEPA: Rethinking the Video JEPA Paradigm for World-Action Modeling in Autonomous Drivingcited by 2
  12. 2026From Representational Complementarity to Dual Systems: Synergizing VLM and Vision-Only Backbones for End-to-End Drivingcited by 1
  13. 2026Driving Intents Amplify Planning-Oriented Reinforcement Learningcited by 1
  14. 2026Causality-Aware End-to-End Autonomous Driving via Ego-Centric Joint Scene Modelingcited by 1
  15. 2026DriveMA: Rethinking Language Interfaces in Driving VLAs with One-Step Meta-Actionscited by 1
  16. 2026NTR: Neural Token Reconstruction for Scene Token Bottleneck in End-to-End Drivingcited by 1
  17. 2026DriveAnchor: Progressive Anchor-based Flow Learning for Autonomous Driving Planningcited by 1
  18. 2026DriveReward: A Comprehensive Dataset and Generative Vision-Language Reward Model for Autonomous Drivingcited by 1
  19. 2026FlowR2A: Learning Reward-to-Action Distribution for Multimodal Driving Planningcited by 1
  20. 2026ASSCG: Just-Right Gating over Chattering for Fast-Slow LLM Planning in Autonomous Drivingcited by 1
  21. 2026UniTeD: Unified Temporal Diffusion for Joint Perception and Planning in Autonomous Drivingcited by 1
  22. 2026Large Discrete Policy: Advancing Explicit Behavior Modeling with Stochastic Iterative Scoringcited by 1
  23. 2026The Constant Eye: Benchmarking and Bridging Appearance Robustness in Autonomous Drivingcited by 0
  24. 2026D3-MoE:Dual Disentangled Diffusion Mixture-of-Experts for Style-Controllable End-to-End Autonomous Drivingcited by 0
  25. 2026MOJITO: Modal Joint Learning for Unified End-to-End Autonomous Drivingcited by 0
  26. 2026Plug-and-Play Traffic Element Awareness for End-to-End Autonomous Drivingcited by 0
  27. 2026Planning in the Backbone: DiffAdapterVLA for Native Continuous Trajectory Generation with Driving VLMscited by 0
  28. 2026MM-Future: Multi-Mode Joint World-Action Modeling for Autonomous Drivingcited by 0
  29. 2026EditWM: Event-Decomposed World Modeling with Incremental Correction for End-to-End Autonomous Drivingcited by 0
  30. 2026Beyond the Leaderboard: Counterfactual Diagnosis of End-to-End and VLA Driving Policies Under Domain Shiftcited by 0
  31. 2026FeasibleFlow: One-Step Joint Transport of Configuration Feasibility and Trajectories for End-to-End Drivingcited by 0
  32. 2026DreamStream: Towards Policy-Oriented Generative Simulation for End-to-End Drivingcited by 0
  33. 2026RECAST: Recasting Vision-Language Semantics into an Actionable Cost Map for Robot Navigationcited by 0
  34. 2026MomWorld: Momentum-Aware Latent World Model for Long-Horizon Autonomous Drivingcited by 0
  35. 2026ReDrive: Shaping Representations with World Modeling for End-to-End Drivingcited by 0
  36. 2026When the Score Becomes the Target: Rethinking Metric Validity in Autonomous Drivingcited by 0
  37. 2026RefineDrive: Reliable Failure-Guided Learning for Vision-Language-Action Drivingcited by 0
  38. 2026World4Scorer: Outcome-Grounded World Modeling for Autonomous Drivingcited by 0

50 most cited of 52 All 52 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2024Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous Driving and Zero-Shot Instruction Following
  2. 2025BridgeDrive: Diffusion Bridge Policy for Closed-Loop Trajectory Planning in Autonomous Driving
  3. 2026Diffusion-2BC: Hybrid Diffusion and Regression Training for Offline Behavior Cloning in Autonomous Driving
  4. 2026Multimodal Action Diffusion for Robust End-to-End Autonomous Driving
  5. 2026Unleashing the Potential of Diffusion Models for End-to-End Autonomous Driving
  6. 2026DiDrive: A Risk-Aware Hierarchical Diffusion Framework for Safe Offline Reinforcement Learning in Autonomous Driving
  7. 2026Fine-tuning is Not Enough: A Parallel Framework for Collaborative Imitation and Reinforcement Learning in End-to-end Autonomous Driving
  8. 2026GSDrive: Reinforcing Driving Policies by Multi-mode Future Trajectory Probing with 3D Gaussian Splatting Environment