Is This the Subspace You Are Looking for? An Interpretability Illusion for Subspace Activation Patching

1 code implementation • 28 Nov 2023 • Aleksandar Makelov, Georg Lange, Neel Nanda

We demonstrate this phenomenon in a distilled mathematical example, in two real-world domains (the indirect object identification task and factual recall), and present evidence for its prevalence in practice.

Attribute

[Re] Reproducibility report of "Interpretable Complex-Valued Neural Networks for Privacy Protection"

1 code implementation • RC 2020 • Arsen Sheverdin, Noud Corten, Alko Knijff, Georg Lange

Since the original paperʼs code was not published, all of the codebase was written independently from scratch, based solemnly on how it was described in the paper.