Research map / 3D vision / 3D generation
HERMES: A Unified Self-Driving World Model for Simultaneous 3D Scene Understanding and Generation
Xin Zhou, Dingkang Liang, Sifan Tu, Xiwu Chen, Yikang Ding, Dingyuan Zhang, Feiyang Tan, Hengshuang Zhao, Xiang Bai
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 |
|---|---|---|---|
| unify understanding and generation in one model instead of separated models sharing BEV (ablation) | HERMES · nuScenes (quarter of training scenes, ablation) | Chamfer distance 3s: 1.875 -> 1.718 | Improved |
| unify understanding and generation instead of separated models (ablation) | HERMES · OmniDrive-nuScenes captioning (quarter data, ablation) | CIDEr: 0.722 -> 0.720 | No clear effect |
| train generation jointly with understanding instead of generation only (ablation) | HERMES · nuScenes (quarter of training scenes, ablation) | Chamfer distance 3s: 1.687 -> 1.718 | Hurt |
| train understanding jointly with generation instead of understanding only (ablation) | HERMES · OmniDrive-nuScenes captioning (quarter data, ablation) | CIDEr: 0.728 -> 0.720 | Hurt |
| add world queries (vs integrating ego motion directly into BEV) (ablation) | HERMES · nuScenes val (full training data) | Chamfer distance 3s: 10% reduction | Improved |
| add world queries (vs none) (ablation) | HERMES · OmniDrive-nuScenes captioning (full data) | CIDEr: 1% decrease | Hurt |
| add current point cloud prediction as auxiliary task (ablation) | HERMES · nuScenes (quarter of training scenes, ablation) | Chamfer distance 3s: 1.779 -> 1.718 | Improved |
| use max pooling instead of attention pooling for world queries (ablation) | HERMES · nuScenes (quarter of training scenes, ablation) | Chamfer distance 3s: 1.748 -> 1.718 | Improved |
| use max pooling instead of average pooling for world queries (ablation) | HERMES · nuScenes (quarter of training scenes, ablation) | Chamfer distance 3s: 1.741 -> 1.718 | Improved |
| increase flattened BEV size from 25x25 to 50x50 (ablation) | HERMES · OmniDrive-nuScenes captioning (quarter data, ablation) | CIDEr: 0.671 -> 0.720 | Improved |
| increase flattened BEV size from 25x25 to 50x50 (ablation) | HERMES · nuScenes (quarter of training scenes, ablation) | Chamfer distance 3s: 1.698 -> 1.718 | Hurt |
| scale LLM from 1.8B to 3.8B (ablation) | HERMES · nuScenes (quarter of training scenes, ablation) | Chamfer distance 3s: 1.718 -> 1.701 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ViDAR — ViDAR: Video Diffusion-Aware 4D Reconstruction From Monocular Inputs (NeurIPS 2025) · better on 3 of 3 shared results
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
- Driving in the Occupancy World: Vision-Centric 4D Occupancy Forecasting and Planning via World Models for Autonomous Driving (AAAI 2025 · similarity 0.88)
- SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving (CVPR 2026 · similarity 0.88)
- Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution (NeurIPS 2025 · similarity 0.88)
- SpaceDrive: Infusing Spatial Awareness into VLM-based Autonomous Driving (CVPR 2026 · similarity 0.87)
- DriveSpatial: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving (NeurIPS 2026 · similarity 0.87)
- Driving into the Future: Multiview Visual Forecasting and Planning with World Model for Autonomous 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.
