Daniel M. Ziegler

Member of Technical Staff at Anthropic

Ziegler led early experiments showing that language models could be trained from human preferences, a direct precursor to reinforcement learning from human feedback. He later worked on practical safety research at Redwood, METR, and Anthropic, connecting alignment methods to tests of increasingly capable models.