Research map / Probabilistic and scientific modeling / Molecules and proteins
MeanFuser: Fast One-Step Multi-Modal Trajectory Generation and Adaptive Reconstruction via MeanFlow for End-to-End Autonomous Driving
Junli Wang, Yinan Zheng, Xueyi Liu, Zebin Xing, Pengfei Li, Kun Ma, Hangjun Ye, Guang Chen, Guang Li, Long Chen, Zhongpu Xia, Qichao Zhang
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 |
|---|---|---|---|
| replace MLP decoder with MeanFlow trajectory generator (ablation) | TransFuser · NAVSIMv1 navtest (ablation) | PDMS: 84.0 -> 87.3 | Improved |
| add Gaussian mixture noise (ablation) | MeanFlow planner · NAVSIMv1 navtest (ablation) | PDMS: 87.3 -> 88.2 | Improved |
| add Adaptive Reconstruction Module instead of classifier scoring (ablation) | MeanFuser · NAVSIMv1 navtest (ablation) | PDMS: 88.2 -> 89.0 | Improved |
| average all proposals instead of using ARM (ablation) | MeanFuser · NAVSIMv1 navtest (ablation) | PDMS: 88.2 -> 71.2 | Hurt |
| add GMN to reduce cases where all proposals leave drivable area (ablation) | MeanFlow planner · NAVSIMv1 navtest (ablation) | N_DAC=0 cases: 143 -> 58 | Improved |
| add ARM to reduce cases where all proposals leave drivable area (ablation) | MeanFuser · NAVSIMv1 navtest (ablation) | N_DAC=0 cases: 58 -> 48 | Improved |
| use 8 Gaussian components instead of 2 (ablation) | MeanFuser · NAVSIMv1 navtest | PDMS: 88.4 -> 89.0 | Improved |
| use 8 Gaussian components instead of 16 (ablation) | MeanFuser · NAVSIMv1 navtest | PDMS: 88.8 -> 89.0 | Improved |
| use 8 Gaussian components instead of 32 (ablation) | MeanFuser · NAVSIMv1 navtest | PDMS: 88.5 -> 89.0 | Improved |
| derive GMN by clustering expert trajectories instead of manual design (ablation) | MeanFuser · NAVSIMv1 navtest | PDMS: 88.6 -> 89.0 | Improved |
| add GMN to MeanFuser (multimodality) (ablation) | MeanFuser · NAVSIMv1 navtest | PDMS: 88.3 -> 89.0 | Improved |
| add GMN to MeanFuser (combined diversity-performance) (ablation) | MeanFuser · NAVSIMv1 navtest | M_DP: 22.07 -> 26.70 | Improved |
Closest papers by meaning
- GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation in End-to-End Autonomous Driving (CVPR 2025 · similarity 0.89)
- ChainFlow-VLA: Causal Flow Planning with Vision-Language Models (NeurIPS 2026 · similarity 0.87)
- Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous and Instruction-guided Driving (CVPR 2024 · similarity 0.87)
- Flow Matching-Based Autonomous Driving Planning with Advanced Interactive Behavior Modeling (NeurIPS 2025 · similarity 0.87)
- SMART: Scalable Multi-agent Real-time Motion Generation via Next-token Prediction (NeurIPS 2024 · similarity 0.86)
- DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving (CVPR 2025 · similarity 0.86)
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.
