lucy3/aftermath_predictions

---
license: cc-by-nc-4.0
---

# Aftermath of DrawEduMath

This contains `predictions.csv`, for recreating the results of the paper titled "The Aftermath of DrawEduMath: Vision Language Models Underperform with Struggling Students and Misdiagnose Errors".

This file contains model predictions for DrawEduMath QA from eleven vision-language models. These models include:
- GPT-4.1
- GPT-4.5 Preview
- o4-mini
- GPT-5
- Claude Sonnet 3.7
- Claude Sonnet 4
- Claude Sonnet 4.5
- Gemini 2.0 Flash
- Gemini 2.5 Pro
- Gemini 2.5 Pro Preview
- Llama 4 Scout

Please consult the datacard for [DrawEduMath](https://huggingface.co/datasets/allenai/DrawEduMath) for detailed information about data source.

Quick links:
- Aftermath of DrawEduMath code: https://github.com/lucy3/aftermath_drawedumath/tree/main
- Aftermath of DrawEduMath paper: https://arxiv.org/abs/2603.00925

## Example Use

```
import pandas as pd

predictions_df = pd.read_csv("predictions.csv")
```

## License

This dataset is licensed under CC-BY-NC-4.0. It is intended for research and educational purposes following ASSISTments's [Responsible Use Guidelines](https://sites.google.com/view/e-trials/resources/guidelines-for-drawedumath).

## Citation

```
@misc{lucy2026aftermathdrawedumathvisionlanguage,
title={The Aftermath of DrawEduMath: Vision Language Models Underperform with Struggling Students and Misdiagnose Errors},
author={Li Lucy and Albert Zhang and Nathan Anderson and Ryan Knight and Kyle Lo},
year={2026},
eprint={2603.00925},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2603.00925},
}
```