question
active
question:how-can-we-scale-head-loadings-and-qk-attributions-to-long-contextsHow can we scale head loadings and QK attributions to long contexts?
Quadratic scaling.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
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.
- Future work.
- Central claim.
- Writes attention scores/head output as sums of feature interactions; core HeadVis capability.
- Preliminary analyses suggest QK attribution matrices per head and position pair are approximately low-rankfinding0.777Motivates distillation.
- In the Aunt Mary Sue prompt, QK attributions matrix has rank at least 2 with no strong cross termsfinding0.754Parallel independent QK interactions.
- Feature attribution (gradient-based) correlates 0.8 with ablation effects on the 'John' and 'Kobe' examples.finding0.748Validation of attribution as a fast proxy for causal importance.
- Table 2; e.g., Identity +1.8 for turn-averaged.