Neurons are predictable. We can generate short, natural language descriptions of neurons' roles. Providing the description to a judge model lets us predict when the neuron will fire, simulate its behavior, and activate it more strongly than anything in our corpus.
Goodfire says natural-language neuron descriptions predict and activate neurons
AIGoodfire says it can generate short natural-language descriptions of neurons' roles. Given to a judge model, the descriptions predict when a neuron will fire and simulate its behavior.
Source details
Goodfire · x.comPublished
Original title: Neurons are predictable. We can generate short, natural language descriptions of neurons' roles. Providing the description to a judge mod...