Pith. sign in

Cross-layer retrospective retrieving via layer attention

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

1 Pith paper citing it

fields

cs.LG 1

years

2025 1

verdicts

CONDITIONAL 1

representative citing papers

You Do Not Fully Utilize Transformer's Representation Capacity

cs.LG · 2025-02-13 · conditional · novelty 5.0

Adding learned per-head routing over previous layers' key-value buffers to a Transformer reduces representation collapse, lowers language modeling loss, and improves synthetic arithmetic and planning accuracy.

citing papers explorer

Showing 1 of 1 citing paper.

  • You Do Not Fully Utilize Transformer's Representation Capacity cs.LG · 2025-02-13 · conditional · none · ref 7

    Adding learned per-head routing over previous layers' key-value buffers to a Transformer reduces representation collapse, lowers language modeling loss, and improves synthetic arithmetic and planning accuracy.