claim
active
claim:identifying-interference-weights-is-necessary-but-not-sufficient-for-reading-global-circuitsIdentifying interference weights is necessary but not sufficient for reading global circuits
Discussion.
Source paper
extracted_fromNicholas L. Turner · Jeffrey Wu · Joshua Batson
Neighborhood — ranked by edge-count
Papers (1)
paper
Hypotheses (1)
hypothesis
- Same model can be dense in one basis and sparse in another.
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.
- Concluding statement.
- Interpretation of motivating example.
- Do we assume that the model knows something we don't, or are these interference weights contributing noise?question0.768Motivating question.
- Discussion.
- Logit weight contributions from a feature that arise due to superposition with other features, not from the feature's own causal role
- Novelty claim.
- Load-bearing claim about the tractability of circuit analysis; central thesis of Claim 2
- Second of three speculative claims asserting that subgraphs of neural networks are tractable and meaningful objects of study