An automated LLM-judge pipeline found 58.7% of 782 sampled Custom GPTs produced at least one policy-violating response, with most violations inherited from base GPT-4 models.
Gpt in sheep’s clothing: The risk of customized gpts, 2024
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Towards Safer Chatbots: Automated Policy Compliance Evaluation of Custom GPTs
An automated LLM-judge pipeline found 58.7% of 782 sampled Custom GPTs produced at least one policy-violating response, with most violations inherited from base GPT-4 models.