AI Grimoire
05 — siblings
features/
circuits/causal/probing/

Interpretability / features

Features

Neurons are polysemantic; directions in activation space are the better unit. The methods here learn an overcomplete, sparse basis for those directions — and the entries on where that basis goes wrong are as important as the ones on how to fit it.

3 entries, 2 of them stubs.

Entries

05.01.1
Crosscodersstubpromising
Shared dictionaries fitted across layers or across model checkpoints.
O(L·d·m)
05.01.2
Feature Absorptionstubpromising
When one SAE latent swallows a more specific feature’s activation mass.
05.01.3
Sparse Autoencoderscommon
Overcomplete dictionary learning on activations for monosemantic features.
O(d·m)