Eric Wallace—Poisoning Language Models During Instruction Tuning

Описание: Eric (ML PhD at Berkeley) works on improving language models, enhancing the security, privacy and robustness of ML, and the intersection of these topics. In this video he presents his poster for his paper "Poisoning Language Models During Instruction Tuning" at the ICML conference.

Paper: https://arxiv.org/pdf/2305.00944.pdf