ComplementaryExpl

Code for the MMR exemplar-selection strategy described in the paper Complementary Explanations for Effective In-Context Learning (ACL Findings, 2023).

Setup

• python==3.8
• requirements: pip install -r requirements.txt
• Set OPENAI KEY: export KEY=yourkey

Experiments

Preparation: sh prepare_data.sh

Example selection exeperiments on GSM:

sh exp_scripts/gsm.sh

Example selection exeperiments ECQA:

sh exp_scripts/ecqa.sh

Example selection exeperiments ESNLI:

sh exp_scripts/esnli.sh

Citation

@InProceedings{Ye-Et-Al:2023:Complementary,
  title = {Complementary Explanations for Effective In-Context Learning},
  author = {Xi Ye and Srinivasan Iyer and Asli Celikyilmaz and Ves Stoyanov and Greg Durrett and Ramakanth Pasunuru},
  booktitle = {Findings of ACL},
  year = {2023},
}