pith. sign in

topic word

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it

fields

cs.LG 1

years

2026 1

verdicts

UNVERDICTED 1

representative citing papers

Focus and Dilution: The Multi-stage Learning Process of Attention

cs.LG · 2026-05-02 · unverdicted · novelty 6.0

In one-layer Transformers trained on Markovian data, attention undergoes a cycle of rapid rank-one condensation, frequency-driven focus on high-frequency tokens, dilution via embedding perturbations, and restart from low-frequency asymmetries.

citing papers explorer

Showing 1 of 1 citing paper.

  • Focus and Dilution: The Multi-stage Learning Process of Attention cs.LG · 2026-05-02 · unverdicted · none · ref 3

    In one-layer Transformers trained on Markovian data, attention undergoes a cycle of rapid rank-one condensation, frequency-driven focus on high-frequency tokens, dilution via embedding perturbations, and restart from low-frequency asymmetries.