method
active
method:toy-model-of-crosscoder-diffing

Toy model of crosscoder diffing

Synthetic two-model activations built from shared and exclusive latent factors with controllable ground truth.

Neighborhood — ranked by edge-count

Papers (1)

paper

Methods (1)

method
  • Train a single SAE encoding and decoding activations from two models simultaneously to find shared and exclusive features.

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.