Papers by Bryan Christ
MATHWELL: Generating Educational Math Word Problems Using Teacher Annotations (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing models and data fail to be educationally appropriate, causing teachers to write boilerplate questions and use boilerplate question sets. |
| Approach: | They propose that large language models (LLMs) can generate educational word problems by generating word problems using annotations from experts. |
| Outcome: | The proposed model generates more solvable, accurate, and appropriate word problems than public models while avoiding harmful questions. |