- Section 4 Human evaluation:
notebooks/analyze_human_and_proxy_results.ipynb
(this notebook also collects the automatic evaluation results) - Section 5 Automatic evaluation:
scripts/proxy_evaluation.py
- Section 6 Strategies
scripts/persuasion_strategy.py
- Section 7 A structural analysis
notebooks/generate_graph_example.ipynb
notebooks/check_proxy_completeness.ipynb
is used to check the completeness of experiment results.scripts/llm.py
contains a tool for prompting LLMs.
@article{ajwani2024generated,
title={{LLM-generated Black-box Explanations can be Adversarially Helpful}},
author={Ajwani, Rohan and Javaji, Shashidhar Reddy and Rudzicz, Frank and Zhu, Zining},
journal={arXiv preprint arXiv:2405.06800},
year={2024}
}