AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Block-Skim: Efficient Question Answering for Transformer

Yue Guan, Zhengyi Li, Zhouhan Lin, Yuhao Zhu, Jingwen Leng, Minyi Guo

AAAI 2022 · Architectures and efficiency › Attention and transformers · approach: Efficient language transformers · 21 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
Jointly train the Block-Skim loss with the QA objectiveBERT-base · SQuAD, HotpotQA, NewsQA, NaturalQuestions, TriviaQA, SearchQAF1: 76.78 -> 77.36Improved
Skim predicted irrelevant blocks during inferenceBERT-base · SQuAD, HotpotQA, NewsQA, NaturalQuestions, TriviaQA, SearchQAF1: 88.32 -> 88.52; 74.39 -> 74.47; 66.57 -> 65.14; 78.85 -> 78.48; 72.61 -> 72.80; 79.93 -> 79.84Mixed
Skim predicted irrelevant blocks during inferenceBERT-base · Six QA datasetsSpeedup: 3.01×; 2.28×; 2.53×; 2.56×; 1.81×; 3.17×Improved
Freeze the Transformer and train only the Block-Skim modules instead of joint training (ablation)Block-Skim · SQuADF1: 88.92 -> 88.32Hurt
Skim blocks during joint QA training (ablation)Block-Skim · SQuADF1: 88.92 -> 86.83Hurt
Use block size 1 instead of 32 (ablation)Block-Skim · SQuADF1: 88.92 -> 88.60Hurt
Use block size 8 instead of 32 (ablation)Block-Skim · SQuADF1: 88.92 -> 88.63Hurt
Use block size 16 instead of 32 (ablation)Block-Skim · SQuADF1: 88.92 -> 88.75Hurt
Use block size 64 instead of 32 (ablation)Block-Skim · SQuADF1: 88.92 -> 88.65Hurt
Use block size 128 instead of 32 (ablation)Block-Skim · SQuADF1: 88.92 -> 88.33Hurt
Add supporting-fact evidence labels to the skim predictor objective (ablation)Block-Skim · HotpotQAF1: 74.88 -> 74.85No clear effect

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