concept
active
concept:safety-relevant-featuresSafety-relevant features
Features that activate on content related to potential harms (deception, bias, dangerous information).
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- Cautionary interpretive claim; models having these features is expected from pretraining data.
- SAEs uncover safety-relevant representations that might be monitored or controlled.
- Scalar weight on loss for each feature.
- The extremely sparse (~3%) set of parameters governing safety alignment, consistent with its ease of disruption by fine-tuning
- Functionally significant directions in the residual stream that organize activations in mechanistic interpretability
- features serve as oversight signals to intentionally design models with desirable capabilitiesquote0.747Core vision statement.
- Internal concepts a model uses, extracted as human-interpretable components.
- Removal of harmful and dead features from released SAEs.