AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Adversarial Self-Attention for Language Understanding

Hongqiu Wu, Ruixue Ding, Hai Zhao, Pengjun Xie, Fei Huang, Min Zhang

AAAI 2023 · Architectures and efficiency › Attention and transformers · approach: Sparse and structured attention · 11 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add ASA fine-tuning to BERTBERT · SST-2accuracy: 93.2 -> 94.1Improved
Add ASA fine-tuning to BERTBERT · MNLIaccuracy: 84.1 -> 85.0Improved
Add ASA fine-tuning to BERTBERT · QNLIaccuracy: 90.4 -> 91.4Improved
Add ASA fine-tuning to RoBERTaRoBERTa · QQPF1: 72.2 -> 73.7Improved
Add ASA fine-tuning to BERTBERT · STS-BSpearman correlation: 84.7 to 86.5Improved
Add ASA fine-tuning to BERTBERT · WNUT-17F1: 47.8 -> 49.8Improved
Add ASA fine-tuning to BERTBERT · DREAMaccuracy: 62.9 to 64.3Improved
Add ASA fine-tuning to BERTBERT · ANLIaccuracy: 48.0 -> 50.4Improved
Add ASA fine-tuning to BERTBERT · PAWS-QQPaccuracy: 81.7 -> 87.7Improved
Add ASA fine-tuning to BERTBERT · HellaSWAGaccuracy: 39.7 -> 40.8Improved
Replace ASA with random Bernoulli attention masking (ablation)ASA · PAWS-QQPaccuracy: 87.7 vs 86.1Hurt
Replace ASA with scheduled Bernoulli masking (ablation)ASA · PAWS-QQPaccuracy: 87.7 vs 85.4Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy