A Study on Question-Answer Dataset for LLM Safety Evaluation with a Focus on Illegal Activities
For researchers and developers working on LLM safety, this provides a dataset and evaluation framework, but the contribution is incremental as it builds on existing work without demonstrating novel impact.
The paper presents a question-answer dataset for evaluating LLM safety regarding illegal activities, based on manual analysis of AnswerCarefully, and introduces creation methods and evaluation rubrics. No concrete results or numbers are reported.
In this paper, we discuss question-answer dataset for LLM safety evaluation, with a focus on illegal activities. Specifically, on the basis of manual analysis of AnswerCarefully, we introduce several additional information, methods for creating question-answer examples, and a rubric for evaluating LLM-generated responses. The outcomes of this study are intended to be shared with the "JAI-Trust" project.