claim
active
claim:holding-the-sae-error-term-fixed-leaves-everything-the-sae-hasn-t-captured-unaltered-by-an-interventionHolding the SAE error term fixed leaves everything the SAE hasn't captured unaltered by an intervention
Design property of the intervention method.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
Methods (1)
method
- Modifying SAE hidden activations h to h-tilde while keeping error term fixed, then continuing inference.
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.
- Explains weaker datamix SAEs.
- Proposed future direction.
- A promising property for interpretability analysis off-distribution.
- Shows gating effect is specific to the self-referential computational regime, not a general feature effect
- Latent #10 activation increase correctly classifies all correct vs incorrect fine-tuned models in Figure 9
- Design decision.
- Surprising finding that the two evaluation methods diverge in their relationship with persistence
- Lower SAE loss at fixed sparsity likely indicates less interference and less superposition.claim0.741Justifies SAE proxy.