framework
active
framework:bayesian-map-interpretation-of-sae-objectiveBayesian MAP interpretation of SAE objective
Lens used to expose SAE priors.
Neighborhood — ranked by edge-count
Papers (1)
paper
Thinkers (1)
thinker
- Sparse coding recovering V1 receptive fields.
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.
- Using GPT-4o or o3 to automatically generate interpretations of SAE latents from top-activating examples
- Proposed future direction.
- Geometric explanation for prior negative results.
- Backprop of logit-minus-mean-logit loss to surface causal intervention candidates, summed across token positions.
- Extension of mechanistic interpretability findings to the metacognitive domain
- Forward-looking claim about the potential of model introspection as an interpretability tool
- Authors' hedged position on SAEs.
- Decomposition approach with rank-1 OV directions.