Research map / Label-efficient and robust learning / Recognition and long-tail
DIVER: Reinforced Diffusion Breaks Imitation Bottlenecks in End-to-End Autonomous Driving
Ziying Song, Lin Liu, Hongyu Pan, Bencheng Liao, Mingzhe Guo, Lei Yang, Yongchang Zhang, Shaoqing Xu, Caiyan Jia, Yadan Luo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add map and agent conditioning to the diffusion generator (ablation) | SparseDrive · Bench2Drive | Driving Score: 44.54 -> 45.34 | Improved |
| add reference GT guidance (ablation) | SparseDrive · Bench2Drive | Driving Score: 44.54 -> 47.12 | Improved |
| add RL with diversity and safety rewards (ablation) | SparseDrive + PADG · Bench2Drive | Driving Score: 47.12 -> 49.21 | Improved |
| add map and agent conditioning (ablation) | TransFuser mmt · NAVSIM v1 | PDMS: 85.1 -> 86.3 | Improved |
| add GRPO RL on top of reference GTs and match loss (ablation) | DIVER PADG · NAVSIM v1 | PDMS: 86.7 -> 88.3 | Improved |
| use GRPO instead of PPO (ablation) | DIVER · NAVSIM v1 | PDMS: 87.9 -> 88.3 | Improved |
| use TrajPooler instead of no image features (ablation) | DIVER · NAVSIM v1 | PDMS: 87.7 -> 88.3 | Improved |
| use TrajPooler instead of full image features (ablation) | DIVER · NAVSIM v1 | PDMS: 87.9 -> 88.3 | Improved |
| add traffic-compliance reward weight 0.1 (ablation) | DIVER · Bench2Drive | Driving Score and Comfortness: 49.21 -> 48.71; 54.72 -> 56.10 | Mixed |
| add PDMS reward weight 0.1 (ablation) | DIVER · NAVSIM v1 | PDMS: 88.3 -> 88.6 | No clear effect |
| apply DIVER to SparseDrive | SparseDrive · nuScenes | avg diversity: 0.13 -> 0.21 | Improved |
| apply DIVER to SparseDrive | SparseDrive · Adv-nuSc | collision rate 3s: 2.430 -> 1.798 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- TransFuser — TransFuser: Imitation With Transformer-Based Sensor Fusion for Autonomous Driving (TPAMI 2022) · better on 2 of 2 shared results
- DriveSuprim — DriveSuprim: Towards Precise Trajectory Selection for End-to-End Planning (AAAI 2026) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous and Instruction-guided Driving (CVPR 2024 · similarity 0.90)
- Diffusion-Based Planning for Autonomous Driving with Flexible Guidance (ICLR 2025 · similarity 0.90)
- DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving (CVPR 2025 · similarity 0.89)
- Model-Based Policy Adaptation for Closed-Loop End-to-end Autonomous Driving (NeurIPS 2025 · similarity 0.88)
- RAD: Training an End-to-End Driving Policy via Large-Scale 3DGS-based Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Beyond Self-Play and Scale: A Behavior Benchmark for Generalization in Autonomous Driving (NeurIPS 2026 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
