Research map / 3D vision / 3D generation
I2-World: Intra-Inter Tokenization for Efficient Dynamic 4D Scene Forecasting
Zhimin Liao, Ping Wei, Ruijie Zhang, Shuaijia Chen, Haoxuan Wang, Ziyang Ren
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 inter-scene tokenization without alignment (ablation) | I2-Scene Tokenizer · Occ3D-nus tokenizer reconstruction ablation | reconstruction mIoU (%): 66.52 -> 70.37 | Improved |
| add ego-pose alignment of historical features in inter-scene tokenization (ablation) | I2-Scene Tokenizer · Occ3D-nus tokenizer reconstruction ablation | reconstruction mIoU (%): 70.37 -> 77.12 | Improved |
| add ego-pose alignment of historical features in inter-scene tokenization (ablation) | I2-Scene Tokenizer · Occ3D-nus tokenizer reconstruction ablation | reconstruction IoU (%): 62.18 -> 64.20 | Improved |
| replace single-scale with multi-scale intra-scene tokenization (ablation) | I2-Scene Tokenizer · Occ3D-nus tokenizer reconstruction ablation | reconstruction mIoU (%): 77.12 -> 81.22 | Improved |
| replace single-scale with multi-scale intra-scene tokenization (ablation) | I2-Scene Tokenizer · Occ3D-nus tokenizer reconstruction ablation | reconstruction IoU (%): 64.20 -> 68.30 | Improved |
| add translation conditioning from transformation matrix (ablation) | I2-Former · Occ3D-nus forecasting ablation | mIoU (%): 17.12 -> 28.74 | Improved |
| add rotation conditioning alone (ablation) | I2-Former · Occ3D-nus forecasting ablation | mIoU (%): 17.12 -> 20.34 | Improved |
| add rotation on top of translation conditioning (ablation) | I2-Former · Occ3D-nus forecasting ablation | mIoU (%): 28.74 -> 34.25 | Improved |
| add lightweight temporal fusion to Inter-Decoder (ablation) | I2-Former · Occ3D-nus forecasting ablation | mIoU (%): 34.25 -> 35.21 | Improved |
| add Intra-Encoder (ablation) | I2-Former · Occ3D-nus forecasting ablation | mIoU (%): 35.21 -> 36.73 | Improved |
| add multi-scale strategy in Intra-Encoder (ablation) | I2-Former · Occ3D-nus forecasting ablation | mIoU (%): 36.73 -> 39.73 | Improved |
| replace Copy-Paste with I2-World (zero-shot, 2Hz) | I2-World · Occ3D-Waymo zero-shot forecasting (2Hz) | mIoU (%): 17.17 -> 36.38 | Improved |
Later papers that compared with this method
2 later accepted papers report I2-World as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Geometry-Aware Spatio-Temporal Context Modeling for 4D Occupancy Forecasting (ECCV 2026 · better on 7 of 7 shared results)
- SparseWorld-TC: Trajectory-Conditioned Sparse Occupancy World Model (CVPR 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
- Semi-Supervised Vision-Centric 3D Occupancy World Model for Autonomous Driving (ICLR 2025 · similarity 0.88)
- GaussianWorld: Gaussian World Model for Streaming 3D Occupancy Prediction (CVPR 2025 · similarity 0.87)
- Driving in the Occupancy World: Vision-Centric 4D Occupancy Forecasting and Planning via World Models for Autonomous Driving (AAAI 2025 · similarity 0.87)
- Semantic Complete Scene Forecasting from a 4D Dynamic Point Cloud Sequence (AAAI 2024 · similarity 0.87)
- X-Scene: Large-Scale Driving Scene Generation with High Fidelity and Flexible Controllability (NeurIPS 2025 · similarity 0.87)
- COME: Adding Scene-Centric Forecasting Control to Occupancy World Model (NeurIPS 2025 · 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.
