New benchmark tests AI safety in Indian languages under persuasive attack

Original: IndicSafeEval: Safety Robustness of Large Language Models under Multilingual Persuasive Jailbreak Attacks

Source: arxiv.org ↗