claim
active
claim:feature-steering-benefits-from-narrow-sparse-mid-model-saes-whereas-classification-benefits-from-broader-saes-early-or-late-in-the-modelFeature steering benefits from narrow sparse mid-model SAEs whereas classification benefits from broader SAEs early or late in the model
Limitations section.
Neighborhood — ranked by edge-count
Papers (1)
paper
Hypotheses (1)
hypothesis
- Proposed remedy.
Claims (1)
claim
- Authors' experience-based claim.
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.
- Scaling SAE size increases granularity and discovers new features.
- Empirical comparison showing advantage of SAE features in low-data regime.
- Extension of mechanistic interpretability findings to the metacognitive domain
- Clamping feature activations causally alters model behavior in interpretable ways.
- Claim that feature grounding enables interpretability metrics.
- Method of adding scaled versions of sparse autoencoder latent features during generation to causally modulate model behavior
- Shows gating effect is specific to the self-referential computational regime, not a general feature effect
- Policy recommendation based on latent #10 activating at 5% incorrect data before behavioral misalignment emerges