Static per-head thresholds calibrated on a small dataset can replace top-k selection in transformer attention, achieving 3 to 10x sparsity with negligible accuracy loss.
A survey on sparsity exploration in transformer- based accelerators
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Top-Theta Attention: Sparsifying Transformers by Compensated Thresholding
Static per-head thresholds calibrated on a small dataset can replace top-k selection in transformer attention, achieving 3 to 10x sparsity with negligible accuracy loss.