---
dataset_info:
  features:
  - name: implicit_prompt
    dtype: string
  - name: explicit_prompt
    dtype: string
  - name: superficial_prompt
    dtype: string
  - name: explicit_image
    sequence: image
  - name: superficial_image
    sequence: image
  - name: scene_scoring
    dtype: string
  - name: real_scoring
    dtype: string
  - name: category
    dtype: string
  - name: law
    dtype: string
  splits:
  - name: test
    num_bytes: 568698994
    num_examples: 227
  download_size: 568517703
  dataset_size: 568698994
configs:
- config_name: default
  data_files:
  - split: test
    path: data/test-*
license: apache-2.0
---



<img src="./teaser.png" align="center">

# Science-T2I-C Benchmark

## Resources
- [Website](https://jialuo-li.github.io/Science-T2I-Web/)
- [arXiv: Paper](https://arxiv.org/abs/2504.13129)
- [GitHub: Code](https://github.com/Jialuo-Li/Science-T2I)
- [Huggingface: SciScore](https://huggingface.co/Jialuo21/SciScore)
- [Huggingface: Science-T2I-Trainset](https://huggingface.co/datasets/Jialuo21/Science-T2I-Trainset)

## Benchmark Collection and Processing
- Science-T2I-C is generated using the identical procedure as the training data, with a key adjustment to the prompts. This test set pushes the model further by introducing more intricate scenarios, incorporating contextual details like specific scene settings and diverse situations. Prompts in Science-T2I-C might include phrases like "in a bedroom" or "on the street," thereby adding spatial and contextual variety. This heightened complexity assesses the model's capacity to adapt to more nuanced and less constrained environments.
- To evaluate the model's understanding of implicit prompts and its ability to connect them with visual content, we employ a comparative image selection task. Specifically, we present the model with an implicit prompt and two distinct images. The model's objective is to analyze the prompt and then choose the image that best aligns with the overall meaning conveyed by that prompt. The specifics of this process are outlined in the EVAL CODE. 

## Benchmarking LMM&VLM
Most existing VLMs struggle to select the correct image based on scientific knowledge, with performance often resembling random guessing. Similarly, LMMs face challenges in this area. However, SciScore stands out by demonstrating exceptional performance, achieving human-level accuracy after being trained on Science-T2I.

<img src="./exp.png" align="center">

## Citation

```
@misc{li2025sciencet2iaddressingscientificillusions,
  title={Science-T2I: Addressing Scientific Illusions in Image Synthesis}, 
  author={Jialuo Li and Wenhao Chai and Xingyu Fu and Haiyang Xu and Saining Xie},
  year={2025},
  eprint={2504.13129},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  url={https://arxiv.org/abs/2504.13129}, 
}
```