Paul Röttger

Examining risks of many-shot jailbreaking in Italian language models.

New dataset reveals challenges in detecting hate speech across languages and platforms.