Max Bileschi
@mlbileschi_pub
Staff Research Software Engineer & Manager at Google Deepmind
ID: 1374745424804843524
24-03-2021 15:30:32
182 Tweet
229 Takipçi
6 Takip Edilen
2+2=5? “LLMs are not Robust to Adversarial Arithmetic” a new paper from our team Google DeepMind with bucket of kets, Laura Culp, AaronParisi, Gamaleldin Elsayed, Jascha Sohl-Dickstein, Noah Fiedel TLDR: We ask an LLM to attack itself and find this works extremely well.