claim
active
claim:per-layer-saes-produce-duplicate-persistent-features-that-make-circuits-seem-more-complex-than-needed-crosscoders-can-radically-simplify-circuitsPer-layer SAEs produce duplicate persistent features that make circuits seem more complex than needed; crosscoders can radically simplify circuits
Motivating example with 13 vs 3 features.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Claims (1)
claim
- Circuit simplification theory.
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.
- Crosscoders are about 2x less FLOP-efficient than per-layer SAEs at equal eval loss at large computefinding0.827Cost of identifying cross-layer structure.
- Indicates redundant structure across layers.
- Decoder cost dominates.
- Crosscoders give a dramatically better whole-model (MSE, L0) tradeoff than summed per-layer SAEsfinding0.799Consolidating shared structure gives concise decomposition.
- Decoder norm profile across 18 layers.
- Motivation for DFC.
- Removing duplicate persistent features to shrink circuit depth.
- Due to SAE training stochasticity.