Record
- Authors
- Heap, T.; Lawson, T.; Farnik, L.; Aitchison, L.
- Year
- 2025
- Title
- Automated Interpretability Metrics Do Not Distinguish Trained and Random Transformers
- Published in
- arXiv
- Type
- Preprint
- DOI
- 10.48550/arxiv.2501.17727
- arXiv
- arXiv:2501.17727
- Our note
- none yet
- Verification
- verified: its identifier resolves to a record with the same title and author (checked against OpenAlex, 2026-11-14).
Identifiers found: OpenAlex W4406975755 · DOI 10.48550/arxiv.2501.17727 · arXiv:2501.17727 - From the lookup
- doi, title, year (the pages' own wording is under "As cited" below)
- Id
heap-2026-automated-interpretability-metrics-do-not-distinguish
Cited on
- Towards Monosemanticity: Decomposing Language Models With Dictionary Learning read status: not stated
As cited
The pages cite this work in one form.
-
Heap, T., Lawson, T., Farnik, L., and Aitchison, L. (2026). Automated Interpretability Metrics Do Not Distinguish Trained and Random Transformers (v2; v1, January 2025, titled "Sparse Autoencoders Can Interpret Randomly Initialized Transformers"). arXiv:2501.17727.
BibLaTeX
@misc{heap-2026-automated-interpretability-metrics-do-not-distinguish,
author = {Heap, T. and Lawson, T. and Farnik, L. and Aitchison, L.},
title = {{Automated Interpretability Metrics Do Not Distinguish Trained and Random Transformers}},
date = {2025},
doi = {10.48550/arxiv.2501.17727},
eprint = {2501.17727},
eprinttype = {arxiv}
}