| # Source and verification
|
|
|
| ## Historical verification (v0.3.0)
|
|
|
| - SAEBase v0.3.0 tag resolves to `d3fd1cb1a453118611f3c7d8b44bd41d1ee7f279`.
|
| - Verified population: 18 configurations and 48 layer sites, each with 64 held-out controlled groups.
|
| - `verified_sites.csv` records the model, SAE, layer, hook, and four measured metrics for each site.
|
| - `metric_results.csv` contains one row for each (site, metric, condition) combination.
|
| - Every record links to a per-site JSON from the source at the pinned commit.
|
| - Numeric records were read from the original source JSON. The metrics were not recomputed.
|
| - Checkpoints are neither included nor redistributed.
|
| - The original raw evaluation remains in separate checksum-verified GitHub releases.
|
|
|
| The v0.3.0 verification establishes engineering acceptance under controlled inputs. It does not establish quality rankings across models, generalization to representative natural-language data, or certification of every checkpoint.
|
|
|
| ## Research Lab measurements (v0.4.0)
|
|
|
| The separately reported 500-prompt Research Lab experiment has its own input population and measurement protocols. See the [v0.4.0 Research Lab report](https://github.com/OpenSAE/SAEBase/blob/main/docs/research-lab/results-20261020.md) for the experiment's source records, limitations, and instructions for verification. The v0.3.0 verification counts above should not be combined with the v0.4.0 research measurements.
|
|
|
