AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Self-Attention Attribution: Interpreting Information Interactions Inside Transformer

Yaru Hao, Li Dong, Furu Wei, Ke Xu

AAAI 2021 · Architectures and efficiency › Attention and transformers · approach: Attention head and block redesign · 186 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

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
Insert attribution-derived triggers into test inputsBERT-base-cased · MNLI entailmentaccuracy: 82.87 -> 0.80Hurt
Insert attribution-derived triggers into test inputsBERT-base-cased · SST-2 positiveaccuracy: 92.79 -> 54.95Hurt
Insert attribution-derived triggers into test inputsBERT-base-cased · MRPC equalaccuracy: 90.32 -> 29.39Hurt
Insert attribution-derived triggers into test inputsBERT-base-cased · RTE entailmentaccuracy: 72.60 -> 9.59Hurt
Rank heads using self-attention attribution scores rather than accuracy differenceBERT-base-cased head pruning · MNLIaccuracyImproved
Retain heads with the highest attribution scoresBERT-base-cased · classification tasksaccuracy: nearly 97%Improved
Use attribution-based head ranking instead of Taylor expansion rankingBERT-base-cased head pruning · MNLIaccuracy: approximately 60%; about 40%Mixed
Use attribution-derived adversarial patternsBERT-based classifiers · MNLI, SST-2, MRPC, and RTEaccuracyHurt

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