My other safety research work:
Explaining success/failure in safety generalization such as zero-shot toxicity reduction (EMNLP 2024 Findings) across languages and finetuning attacks on multilingual LLMs (preprint).
Performing safety red-teaming for frontier NLP/Speech models: Aya (⭑Best Paper Award, ACL 2024), MMS (to appear).
I also worked on making LLMs overcome language barriers and support underrepresented languages. I’ve worked on adapting LLMs to low-resource languages (ACL 2023), generating synthetic data for extremely low-resource languages (EMNLP 2024 Findings), and developing massively multilingual models at frontier AI groups such as Meta AI FAIR (MMS model), Cohere For AI (Aya model) and BigScience (T0, BLOOM, mT0/BLOOMZ, BLOOM+1).