claim
active
claim:qk-attributions-are-a-significant-qualitative-improvement-on-original-attribution-graphs-unlocking-previously-impossible-analysesQK attributions are a significant qualitative improvement on original attribution graphs, unlocking previously impossible analyses
Central claim.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Transcoder-based attribution graphs omit attentional computation, rendering them useless for many promptsextendssupportsMotivating problem.
Questions (1)
question
- Softmax competition problem.
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.
- Core method decomposing attention scores as bilinear sums of query-key feature pairs.
- Writes attention scores/head output as sums of feature interactions; core HeadVis capability.
- Future work.
- Quantifies method limitations.
- Preliminary analyses suggest QK attribution matrices per head and position pair are approximately low-rankfinding0.794Motivates distillation.
- Quadratic scaling.
- QK attribution from Ivory to 'a' has top pair Ivory Coast query feature and Zambian key featurefinding0.781Explains attention edge.
- Feature attribution (gradient-based) correlates 0.8 with ablation effects on the 'John' and 'Kobe' examples.finding0.765Validation of attribution as a fast proxy for causal importance.