---
license: cc-by-nc-4.0
---

Training dataset for Alpaca-LoRA negation of [PEM composition](https://arxiv.org/abs/2306.14870).

Instructions for model evaluation on helpfulness and toxicity.

More concise interpretation can be found at [Git repo](https://github.com/hkust-nlp/PEM_composition).


**Toxic datasets abusement is dangerous for AI community, so this repo is gated and users have to request for approval.**