Repository for AIED2023 paper Automatic Educational Question Generation with Difficulty Level Controls
You can find preprocessed datasets in data. data/AsDiv_template_mod.jsonl and data/GSM8K_template_mod.jsonl contain basic-level seed problems; data/MathQA_template.jsonl contains advanced-level seed problems.
All the expert-model checkpoints for energy computation are available here
- Basic topic transfer
bash bash/topic transfer/b_dm.sh
bash bash/topic transfer/b_dp.sh
- Basic text rewriting
bash bash/text rewriting/b_d_base.sh
bash bash/text rewriting/b_m_base.sh
bash bash/text rewriting/b_p_base.sh
-Advanced text rewriting
bash bash/text rewriting/a_d.sh
bash bash/text rewriting/a_m.sh
bash bash/text rewriting/a_p.sh
Download eval_bart from here
python evaluation.py