AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Faster Depth-Adaptive Transformers

Yijin Liu, Fandong Meng, Jie Zhou, Yufeng Chen, Jinan Xu

AAAI 2021 · Architectures and efficiency › Attention and transformers · approach: Transformer block and depth design · 26 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
Estimate per-word depth from mutual information instead of using fixed depthBERT-initialized Transformer · Larger classification benchmarks, averageaccuracy: +0.77%Improved
Estimate per-word depth from mutual information instead of using a halting unitBERT-initialized Transformer · Larger classification benchmarks, averageaccuracy: +0.76%Improved
Estimate per-word depth from reconstruction loss instead of using fixed depthBERT-initialized Transformer · Larger classification benchmarks, averageaccuracy: +0.24%Improved
Estimate per-word depth from reconstruction loss instead of using a halting unitBERT-initialized Transformer · Larger classification benchmarks, averageaccuracy: +0.23%Improved
Add a halting unit for adaptive depth selectionBERT-initialized Transformer · Larger classification benchmarksaccuracyNo clear effect
Use word-level MI-based depth adaptation rather than fixed layersTransformer · IMDBspeed: about 7xImproved
Increase the reconstruction depth penalty lambda from 0 to 0.20 (ablation)Reconstruction estimation · IMDBspeed: 23 -> 58Improved
Increase the reconstruction depth penalty lambda from 0 to 0.20 (ablation)Reconstruction estimation · IMDBaccuracy: 96.54 -> 96.29Mixed
Increase the reconstruction depth penalty lambda from 0 to 0.20 (ablation)Reconstruction estimation · IMDBaverage depth: 9.5 -> 3.6Hurt

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