Top-k attention sparsity during fine-tuning gave DistilBERT a 0.97% accuracy gain on SST-2, but the result comes from single runs without error bars.
Sparse Transformer: Concentrated Attention Through Explicit Selection
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
method 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
REJECT 1roles
method 1polarities
use method 1representative citing papers
citing papers explorer
-
Crisp Attention: Regularizing Transformers via Structured Sparsity
Top-k attention sparsity during fine-tuning gave DistilBERT a 0.97% accuracy gain on SST-2, but the result comes from single runs without error bars.