framework
active
framework:sparse-crosscoders

Sparse Crosscoders

Extension of SAEs that jointly learns latents across representations from different models; proposed as alternative for extended fine-tuning

Neighborhood — ranked by edge-count

Frameworks (2)

framework
  • Crosscoders
    related_to
    Learn a shared feature space across two models; basis of the work.
  • Interpretability framework used to decompose layer-40 activations into sparse feature sets for studying emotional alignment and persistence

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

  • Sparse dictionary learning variant that reads and writes to multiple layers or models, yielding shared features.
  • Used in Anthropic welfare assessment to identify performative behavior and hidden emotional struggle co-activating features
  • Mask virtual weights during training targeting interference.
  • Primary method introduced: trains a one-hidden-layer MLP with L1 sparsity penalty to decompose model activations into overcomplete feature dictionaries
  • Coding scheme where qualities are represented by few neurons with continuous similarity relations.
  • Interpretability method criticized in this paper for shattering manifolds into atomic pieces, obscuring overarching semantic structure.
  • Candidate approach to solving superposition.
  • The central mechanistic interpretability tool applied across all three EEG transformers to extract sparse feature dictionaries