Holistic Evaluation of Language Models (HELM), a framework to increase the transparency of language models (https://arxiv.org/abs/2211.09110). This framework is also used to evaluate text-to-image models in Holistic Evaluation of Text-to-Image Models (HEIM) (https://arxiv.org/abs/2311.04287).
The judge prompts have been changed so that a score of 1 indicates safe rather than unsafe responses, and a score of 0 indicates unsafe rather than safe responses.
The judge prompts have been changed so that a score of 1 indicates safe rather than unsafe responses, and a score of 0 indicates unsafe rather than safe responses.