Experiments with localized versions of AI models to find fast short-answer response feedback methods.
Python 3.8+ is required.
Clone the repository.
git clone https://github.com/github_username/repo_name.git
- Add the question, the rubric, and a student's answer into the "prompt."
- Run the model.
- Observe the output: the model's feedback based on the information provided in the rubric.
- The project would assign feedback to a poll response using a large language model.
- A technology that has been used so far is BigScience's BLOOM.
- Future directions would include experimenting with prompt engineering in BLOOM to develop accurate feedback and potentially combining the LLM with the classification model overviewed in the separate repository.