claim
active
claim:the-best-way-to-measure-sae-quality-is-a-range-of-diverse-metrics-on-properties-practitioners-care-aboutThe best way to measure SAE quality is a range of diverse metrics on properties practitioners care about
Design philosophy.
Source paper
extracted_from(2025) · Adam Karvonen · Can Rager · Johnny Lin · Curt Tigges +11
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
Questions (1)
question
- Opening question.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Open issue from Pythia vs Gemma.
- Log-loss recovery ratio between SAE-inserted and ablated model.
- Prediction following diversity-undertraining logic.
- Motivates multi-metric evaluation.
- App. B.
- Normative criterion for artificial intelligence derived from variational free energy principle
- Extension of mechanistic interpretability findings to the metacognitive domain
- Policy recommendation based on latent #10 activating at 5% incorrect data before behavioral misalignment emerges