Research map / Vision-language and multimodal / Robot action and navigation
iWorld-Bench: A Benchmark for Interactive World Models with a Unified Action Generation Framework
Jianjie Fang, Yingshan Lei, Qin Wan, Ziyou Wang, Yuchao Huang, Yongyan Xu, Baining Zhao, Weichen Zhang, Chen Gao, Xinlei Chen, Yong Li
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 |
|---|---|---|---|
| use one-hot discrete action control (HY-World 1.5) rather than text control (CogVideoX-I2V) | interactive world models · iWorld-Bench | Trajectory Accuracy: 0.5950 -> 0.7472 | Improved |
| fine-tune HunyuanVideo-1.5 for camera control (HY-World 1.5) | HunyuanVideo-1.5 · iWorld-Bench | Avg score: 0.7188 -> 0.7873 | Improved |
| fine-tune HunyuanVideo-1.5 for camera control (HY-World 1.5) | HunyuanVideo-1.5 · iWorld-Bench | Image Quality: 0.7128 -> 0.6675 | Hurt |
| fine-tune CogVideoX-I2V for camera control (AC3D) | CogVideoX-I2V · iWorld-Bench | Memory Symmetry: 0.6010 -> 0.9068 | Improved |
| fine-tune CogVideoX-I2V for camera control (AC3D) | CogVideoX-I2V · iWorld-Bench | Image Quality: 0.6521 -> 0.4573 | Hurt |
| fine-tune CogVideoX-I2V for camera control (AC3D) | CogVideoX-I2V · iWorld-Bench | Trajectory Accuracy: 0.5950 -> 0.5785 | Hurt |
| compare text-controlled models with other control paradigms | text-controlled world models · iWorld-Bench | Brightness Consistency vs Trajectory Accuracy: 0.8988 brightness consistency vs 0.5950 trajectory accuracy (CogVideoX-I2V) | Mixed |
| use early camera-parameter methods (MotionCtrl, CameraCtrl) instead of recent ones | camera-parameter-controlled models · iWorld-Bench | Avg score: MotionCtrl 0.5486 vs videox-fun-Wan 0.7474 | Hurt |
Closest papers by meaning
- WorldModelBench: Judging Video Generation Models As World Models (NeurIPS 2025 · similarity 0.89)
- Trimming the Long-Tail of Visual World Modeling Evaluation (NeurIPS 2026 · similarity 0.88)
- WorldSimBench: Towards Video Generation Models as World Simulators (ICML 2025 · similarity 0.88)
- LTD-Bench: Evaluating Large Language Models by Letting Them Draw (NeurIPS 2025 · similarity 0.88)
- 4DWorldBench: A Comprehensive Evaluation Framework for 3D/4D World Generation Models (CVPR 2026 · similarity 0.88)
- VS-Bench: Evaluating VLMs for Strategic Abilities in Multi-Agent Environments (CVPR 2026 · 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.
