README.md

---
task_categories:
- any-to-any
language:
- en
configs:
- config_name: main
  data_files:
  - split: ballgame
    path: ballgame/*.parquet
  - split: cube
    path: cube/*.parquet
  - split: maze
    path: maze/*.parquet
  - split: mmsi
    path: mmsi/*.parquet
  - split: multihop
    path: multihop/*.parquet
  - split: paperfolding
    path: paperfolding/*.parquet
  - split: sokoban
    path: sokoban/*.parquet
tags:
- multimodal
- reasoning
- world-models
---

# VisWorld-Eval: Task Suite for Reasoning with Visual World Modeling 🌏

[![Project Page](https://img.shields.io/badge/Project_Page-blue)](https://thuml.github.io/Reasoning-Visual-World/)
[![Paper](https://img.shields.io/badge/arXiv-Paper-b31b1b.svg?logo=arxiv)](https://arxiv.org/abs/2601.19834)
[![GitHub Repo](https://img.shields.io/badge/GitHub-Code-black?logo=github)](https://github.com/thuml/Reasoning-Visual-World)
[![Hugging Face](https://img.shields.io/badge/Hugging_Face-VisWorld_Eval-F8D44E.svg?logo=huggingface)](https://huggingface.co/datasets/thuml/VisWorld-Eval)

![](assets/benchmark.png)

## 📋 Introduction

The VisWorld-Eval suite is for assessing multimodal reasoning with visual world modeling. It comprises seven tasks spanning both synthetic and real-world domains, each designed to isolate and demand specific atomic world-model capabilities.

| Task                         | Capability      | Domain      | Test Samples | Source / Reference              |
|------------------------------|-----------------|-------------|--------------|--------------------------------|
| Paper folding                | Simulation      | Synthetic   | 480          | [SpatialViz](https://github.com/wangst0181/Spatial-Visualization-Benchmark)              |
| Multi-hop manipulation       | Simulation      | Synthetic   | 480          | [ZebraCoT](https://github.com/multimodal-reasoning-lab/Bagel-Zebra-CoT), [CLEVR](https://github.com/facebookresearch/clevr-dataset-gen)    |
| Ball tracking                | Simulation      | Synthetic   | 1,024        | [RBench-V](https://github.com/CHEN-Xinsheng/VLMEvalKit_RBench-V)          |
| Maze                          | Simulation      | Synthetic   | 480          | [maze-dataset](https://github.com/understanding-search/maze-dataset)          |
| Sokoban                       | Simulation      | Synthetic   | 480          | [Game-RL](https://github.com/tongjingqi/Game-RL)                 |
| Cube 3-view projection        | Reconstruction  | Synthetic   | 480          | [SpatialViz](https://github.com/wangst0181/Spatial-Visualization-Benchmark)              |
| Real-world spatial reasoning | Reconstruction  | Real-world  | 522          | [MMSI-Bench](https://github.com/InternRobotics/MMSI-Bench)             |

## ⚙️ Load Data

Load from 🤗 HuggingFace:

```python
from datasets import load_dataset
ds = load_dataset("thuml/VisWorld-Eval")
```

## 🧪 Evaluation

We evaluate different models through API servers. Set your API key and server address in `evaluate.py` (for the evaluated model) and `task_verifiers.py` (for the model judge), then run the evaluation script.

```bash
python eval/evaluate.py --task ballgame --model gemini3pro
```

## 🏆 Leaderboard

Zero-shot evaluation of advanced VLMs on VisWorld-Eval: We report the average accuracy over five tasks (excluding Maze and Sokoban) and over all seven tasks.

| Models                 | Paper Folding | Multi-Hop Manip. | Ball Tracking | Cube 3-View | MMSI (Pos. Rel.) | Maze | Sokoban | Overall (5 tasks) | Overall (7 tasks) |
|------------------------|---------------|------------------|---------------|------------|------------------|------|---------|-------------------|-------------------|
| Gemini 3 Flash         | 25.6          | **75.4**             | **55.3**          | 52.7       | 41.3             | 73.9 | **99.3**    | **50.0**              | **60.5**              |
| Gemini 3 Pro           | **27.0**          | 74.5             | 44.7          | **53.3**       | **49.6**             | 33.5 | 90.2    | 49.8              | 53.2              |
| Seed 1.8               | 10.6          | 75.2             | 24.4          | 42.5       | 38.8             | **83.9** | 68.3    | 38.3              | 49.1              |
| GPT 5.1                | 6.4           | 73.9             | 34.8          | 44.5       | 44.8             | 0.6  | 62.8    | 40.8              | 38.2              |
| o3                     | 13.5          | 68.1             | 24.7          | 37.7       | 44.4             | 0.0  | 36.0    | 37.6              | 32.0              |
| Qwen3-VL-8B-Thinking    | 11.0          | 49.3             | 17.8          | 21.2       | 27.7             | 0.0  | 5.8     | 25.4              | 18.9              |
| BAGEL-7B-MoT           | 11.2          | 31.6             | 19.4          | 26.8       | 27.2             | 0.0  | 0.2     | 23.2              | 16.6              |

## 🚀 Release Progress

- [x] VisWorld-Eval data
- [x] VisWorld-Eval evaluation scripts

## 📜 Citation

If you find this project useful, please cite our paper as:

```
@article{wu2026visual,
    title={Visual Generation Unlocks Human-Like Reasoning through Multimodal World Models}, 
    author={Jialong Wu and Xiaoying Zhang and Hongyi Yuan and Xiangcheng Zhang and Tianhao
Huang and Changjing He and Chaoyi Deng and Renrui Zhang and Youbin Wu and Mingsheng
Long},
    journal={arXiv preprint arXiv:2601.19834},
    year={2026},
}
```

## 🤝 Contact

If you have any questions, please contact wujialong0229@gmail.com.

## 💡 Acknowledgement

We sincerely appreciate the following projects for their valuable codebase and task design: [SpatialViz](https://github.com/wangst0181/Spatial-Visualization-Benchmark), [RBench-V](https://github.com/CHEN-Xinsheng/VLMEvalKit_RBench-V), [maze-dataset](https://github.com/understanding-search/maze-dataset), [Game-RL](https://github.com/tongjingqi/Game-RL), [clevr-dataset-gen](https://github.com/facebookresearch/clevr-dataset-gen), [MMSI-Bench](https://github.com/InternRobotics/MMSI-Bench).