Research map / Video / Video generation
MicroVerse: A Preliminary Exploration Toward a Micro-World Simulation
Rongsheng Wang, Minghao Wu, Hongru Zhou, Zhihan Yu, Zhenyang Cai, Junying Chen, Benyou Wang
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 |
|---|---|---|---|
| fine-tune on MicroSim-10K | Wan2.1-T2V-1.3B · MicroWorldBench | scientific fidelity: 40.3 -> 43.0 | Improved |
| fine-tune on MicroSim-10K | Wan2.1-T2V-1.3B · MicroWorldBench | visual quality: 71.8 -> 68.5 | Hurt |
| fine-tune on MicroSim-10K | Wan2.1-T2V-1.3B · MicroWorldBench | instruction following: 50.1 -> 49.3 | Hurt |
| fine-tune on MicroSim-10K | Wan2.1-T2V-1.3B · MicroWorldBench subcellular-level | rubric score: 52.4 -> 53.3 | Improved |
| add general-domain data to MicroSim-10K (mixed-domain training) (ablation) | MicroVerse 1.3B · MicroWorldBench | scientific fidelity: 43.0 -> 44.1 | Improved |
| add general-domain data to MicroSim-10K (mixed-domain training) (ablation) | MicroVerse 1.3B · MicroWorldBench | visual quality: 68.5 -> 74.4 | Improved |
| add general-domain data to MicroSim-10K (mixed-domain training) (ablation) | MicroVerse 14B · MicroWorldBench | scientific fidelity: 45.4 -> 48.3 | Improved |
| fine-tune 14B only on MicroSim-10K | Wan2.1-14B · MicroWorldBench | visual quality: 86.0 -> 82.7 | Hurt |
| fine-tune 14B only on MicroSim-10K | Wan2.1-14B · MicroWorldBench | scientific fidelity: 42.7 -> 45.4 | Improved |
| scale model from 1.3B to 14B (baseline Wan2.1) | Wan2.1 · MicroWorldBench | scientific fidelity: 40.3 -> 42.7 | No clear effect |
| scale model from 1.3B to 14B (baseline Wan2.1) | Wan2.1 · MicroWorldBench | visual quality: 71.8 -> 86.0 | Improved |
| use expert-filtered MicroSim-10K instead of raw uncleaned data (ablation) | MicroVerse 1.3B · MicroWorldBench | instruction following: 44.5 -> 49.3 | 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.
- CogVideoX-5B — CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer (ICLR 2025) · better on 2 of 2 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
- PhysGen3D: Crafting a Miniature Interactive World from a Single Image (CVPR 2025 · similarity 0.83)
- Micro-Bench: A Microscopy Benchmark for Vision-Language Understanding (NeurIPS 2024 · similarity 0.83)
- How Far Is Video Generation from World Model: A Physical Law Perspective (ICML 2025 · similarity 0.83)
- WISA: World simulator assistant for physics-aware text-to-video generation (NeurIPS 2025 · similarity 0.83)
- PhysCtrl: Generative Physics for Controllable and Physics-Grounded Video Generation (NeurIPS 2025 · similarity 0.83)
- VideoPhy: Evaluating Physical Commonsense for Video Generation (ICLR 2025 · similarity 0.82)
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.
