Research map / Video / Video generation
IC-World: In-Context Generation for Shared World Modeling
FAN WU, Jiacheng Wei, Ruibo Li, Yi Xu, junyou li, Deheng Ye, Guosheng Lin
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 in-context generation (grid prompt) zero-shot (ablation) | LTX-Video-13B · Static scene + dynamic camera | M-FID: 115.3138 -> 94.1303 | Improved |
| add in-context generation (grid prompt) zero-shot (ablation) | LTX-Video-13B · Dynamic scene + static camera | M-FID: 200.0291 -> 235.7100 | Hurt |
| add in-context generation (grid prompt) zero-shot (ablation) | LTX-Video-13B · Dynamic scene + static camera | Motion_20: 0.8081 -> 0.8265 | Improved |
| add in-context generation (grid prompt) zero-shot (ablation) | Wan2.1-14B-distill · Static scene + dynamic camera | VLM score: 0.7193 -> 0.7818 | Improved |
| add in-context generation (grid prompt) zero-shot (ablation) | Wan2.1-14B-distill · Dynamic scene + static camera | M-FID: 185.2354 -> 160.0052 | Improved |
| add in-context generation (grid prompt) zero-shot (ablation) | Wan2.1-14B-distill · Static scene + dynamic camera | Geometry_0.7: 0.6818 -> 0.7006 | Improved |
| use GRPO instead of SFT (1K data) (ablation) | LoRA on Wan2.1-14B-distill · Static scene + dynamic camera | Geometry_0.5: 0.7079 -> 0.7217 | Improved |
| double SFT data from 1K to 2K (ablation) | LoRA SFT on Wan2.1-14B-distill · Static scene + dynamic camera | Geometry_0.5: 0.7079 -> 0.7121 | Improved |
| add SFT warm-up before GRPO (ablation) | LoRA GRPO on Wan2.1-14B-distill · Static scene + dynamic camera | Geometry_0.5: 0.7217 -> 0.7193 | No clear effect |
| add geometry reward (ablation) | IC-World GRPO · Static scene + dynamic camera | Geometry_0.5: 0.7180 -> 0.7270 | Improved |
| add motion reward (ablation) | IC-World GRPO · Dynamic scene + static camera | Motion_20: 0.8364 -> 0.8445 | Improved |
| add both geometry and motion rewards (ablation) | IC-World GRPO · Dynamic scene + static camera | M-FID: 165.4712 -> 157.6099 | Improved |
Closest papers by meaning
- Learning World Models for Interactive Video Generation (NeurIPS 2025 · similarity 0.88)
- OptiWorld: Optimal Control for Video World Generation under Physical Constraints (NeurIPS 2026 · similarity 0.87)
- WorldWeaver: Generating Long-Horizon Video Worlds via Rich Perception (NeurIPS 2025 · similarity 0.87)
- WorldModelBench: Judging Video Generation Models As World Models (NeurIPS 2025 · similarity 0.87)
- Video World Models with Long-term Spatial Memory (NeurIPS 2025 · similarity 0.87)
- LongScape: Advancing Long-Horizon Embodied World Models with Context-Aware MoE (NeurIPS 2026 · 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.
