Kirill Acharya

@kirillacharya

PhD student @Stanford

following
followers

Matryoshka Attribution debuts with a claimed No. 1 result on the Mechanistic Interpretability Benchmark

The paper’s authors say MAttr uses gradient descent to find which parts of a neural network are responsible for a behavior. They report a benchmark result of 2.9× the runner-up.

9 Sources, 22d ago, updated 22d ago

Featured in

Matryoshka Attribution debuts with a claimed No. 1 result on the Mechanistic Interpretability Benchmark

Featured in

Matryoshka Attribution debuts with a claimed No. 1 result on the Mechanistic Interpretability Benchmark