concept
active
concept:attention-headsAttention heads
Transformer attention heads that could be recruited to extract different kinds of information (text vs. thoughts).
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Attention computations distribute across heads via parameter subcomponents with interpretable rolescitesMechanistic discovery about how attention mechanisms decompose into interpretable parameter components.
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.
- Analysis measuring whether each attention head's maximum attention increase points to the correct injected sentence
- A transformer variant where OV and QK matrices of different attention heads can share components, enabling shared copying mechanisms
- The composition of two attention heads via V-composition, forming a new entity with its own attention pattern A^h2 * A^h1 and OV matrix W_OV^h2 * W_OV^h1
- A form of key-query attention within a single input sequence; core to Transformers.
- Process using Q, K, V to compute a heat map over K and weighted sum of V.
- A predictive model representing and controlling attention; central to attention schema theory.
- Core operation in transformers, computing weighted combinations of previous elements
- Mathematical equivalence enabling independent analysis of each attention head