Aymara AI evaluates 20 commercial LLMs across 10 safety domains via policy-grounded adversarial prompts, reporting mean safety scores from 86.2% to 52.4% with large domain-specific gaps.
Manning, Christopher R´ e, Diana Acosta-Navas, Drew A
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Policy-Grounded Safety Evaluation of 20 Large Language Models
Aymara AI evaluates 20 commercial LLMs across 10 safety domains via policy-grounded adversarial prompts, reporting mean safety scores from 86.2% to 52.4% with large domain-specific gaps.