A′ = A(n∈{2,3}) ∨ Ā(n=3) — B3/S23 · Conway, 1970
Attention(Q,K,V) = softmax(QKᵀ/√d
k
)V
F = S + A + C — Reynolds, 1986
db
lab
db
lab
db/ux
redesigning the web for agents