pith. sign in

Gqa: Training generalized multi-query transformer models from multi-head checkpoints

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

1 Pith paper citing it

fields

cs.LG 1

years

2026 1

verdicts

ACCEPT 1

representative citing papers

Stochastic Sparse Attention for Memory-Bound Inference

cs.LG · 2026-05-03 · accept · novelty 6.0

SANTA sparsifies post-softmax value aggregation via stratified sampling of S << n_k indices to produce an unbiased estimator, delivering 1.5x decode attention speedup on RTX 6000 Ada at 32k contexts while matching baseline accuracy.

citing papers explorer

Showing 1 of 1 citing paper.

  • Stochastic Sparse Attention for Memory-Bound Inference cs.LG · 2026-05-03 · accept · none · ref 1

    SANTA sparsifies post-softmax value aggregation via stratified sampling of S << n_k indices to produce an unbiased estimator, delivering 1.5x decode attention speedup on RTX 6000 Ada at 32k contexts while matching baseline accuracy.