Research map / Reinforcement learning / Visual and representation RL
ReconDreamer-RL: Enhancing Reinforcement Learning via Diffusion-based Reconstruction
Chaojun Ni, Guosheng Zhao, Xiaofeng Wang, Zheng Zhu, Wenkang Qin, Chen Xinze, Guanghong Jia, Guan Huang, Wenjun Mei
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 video diffusion prior to ReconSimulator (ablation) | ReconDreamer-RL · nuScenes novel-view rendering, lane shift 6m | FID: 256.42 -> 140.24 | Improved |
| add video diffusion prior to ReconSimulator (ablation) | ReconDreamer-RL · nuScenes novel-view rendering, lane shift 6m | NTA-IoU: 0.148 -> 0.302 | Improved |
| add video diffusion prior to ReconSimulator (ablation) | ReconDreamer-RL · nuScenes novel-view rendering, lane change | FID: 178.34 -> 124.31 | Improved |
| add video diffusion prior to ReconSimulator (ablation) | ReconDreamer-RL · Waymo novel-view rendering, lane change | FID: 89.83 -> 72.86 | Improved |
| add video diffusion prior to ReconSimulator (ablation) | ReconDreamer-RL · Waymo novel-view rendering, lane change | NTA-IoU: 0.486 -> 0.552 | Improved |
| add video diffusion prior to ReconSimulator (ablation) | ReconDreamer-RL · nuScenes novel-view rendering, lane shift 6m | NTL-IoU: 44.89 -> 48.94 | Improved |
| add ReconSimulator to training (ablation) | ReconDreamer-RL · nuScenes closed-loop 3DGS benchmark (ablation) | Collision Ratio: 0.238 -> 0.172 | Improved |
| add DAA on top of ReconSimulator (ablation) | ReconDreamer-RL · nuScenes closed-loop 3DGS benchmark (ablation) | Collision Ratio: 0.172 -> 0.117 | Improved |
| add CTG on top of ReconSimulator (ablation) | ReconDreamer-RL · nuScenes closed-loop 3DGS benchmark (ablation) | Collision Ratio: 0.172 -> 0.143 | Improved |
| add CTG on top of ReconSimulator (ablation) | ReconDreamer-RL · nuScenes closed-loop 3DGS benchmark (ablation) | Deviation Ratio: 0.073 -> 0.053 | Improved |
| add ReconSimulator to DAA + CTG (ablation) | ReconDreamer-RL · nuScenes closed-loop 3DGS benchmark (ablation) | Collision Ratio: 0.122 -> 0.077 | Improved |
| add DAA to ReconSimulator + CTG (ablation) | ReconDreamer-RL · nuScenes closed-loop 3DGS benchmark (ablation) | Collision Ratio: 0.143 -> 0.077 | Improved |
Closest papers by meaning
- ReconDreamer: Crafting World Models for Driving Scene Reconstruction via Online Restoration (CVPR 2025 · similarity 0.91)
- RAD: Training an End-to-End Driving Policy via Large-Scale 3DGS-based Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- RLGF: Reinforcement Learning with Geometric Feedback for Autonomous Driving Video Generation (NeurIPS 2025 · similarity 0.88)
- DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving (CVPR 2025 · similarity 0.87)
- ReSim: Reliable World Simulation for Autonomous Driving (NeurIPS 2025 · similarity 0.87)
- Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous and Instruction-guided Driving (CVPR 2024 · similarity 0.87)
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.
