trends.barnas.net

method · first seen 30 Oct 2026

attribution graph

A graph of causally connected interpretable features that decomposes a language model's computation for analysis.

Mentions per week

0219 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

transcodersparse autoencoderTopK sparse autoencoder