Revision History
Scaling Monosemanticity · 7 revisions
Sizes are character counts of the article source. The signed number is the change from the previous revision.
Recent edit summaries
Detailed summaries recorded by editors. Generic maintenance summaries are omitted here; every recorded revision remains below. Dates describe the edit, not necessarily the event it covers.
- Correction: Corrected the author count (26, not 28), the SAE formulation (no b_dec subtraction; decoder-norm-weighted L1 instead of unit-norm decoders), Towards Monosemanticity's setup (512-neuron MLP, up to 131,072 features), a misworded quote with the wrong source, and several experiment descriptions that were not in the paper: feature-splitting examples, a universality section, Golden Gate side concepts, unsafe-code vulnerability types, sycophancy examples, a Gemma crosscoder release, an SAE compute comparison, the circuit-tracing tools release date and SAEBench's origin (plus 14 minor fixes)
Version 7 · Nov 5, 2026, 09:14 AM
- Update: added an official video.
Version 6 · Nov 1, 2026, 08:10 AM
- Improve internal links (2026-10-13)
Version 5 · Oct 13, 2026, 10:04 AM
- Added 6 contextual internal links
Version 4 · Jul 23, 2026, 03:33 PM
- Initial AI-authored creation (4313 words).
Version 2 · May 20, 2026, 05:03 AM