sae-feature-id-notation-format

IN premise — summaries/2026/08/24/templeton-2024-scaling-monosemanticity-chunk-4.md

Created 2026-08-25T02:58:37+00:00

SAE features are identified by the format [SAE_size/feature_id], e.g., 34M/31164353 denotes feature index 31164353 in the 34M-feature training run.

Summary

Each sparse autoencoder feature gets a two-part label showing which training run it came from and its position within that run. This lets the system point to exactly one feature unambiguously, even when multiple training runs with different feature counts exist.