Anthropic researchers have used a tool called the 'Jacobian Lens' to gain an unprecedented glimpse into the inner workings of a large language model.

Source Brief

Anthropic researchers have used a tool called the 'Jacobian Lens' to gain an unprecedented glimpse into the inner workings of a large language model. They observed Claude 'puzzling over' concepts—a step towards understanding the often opaque 'black box' of AI.