AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Understanding and Improving Encoder Layer Fusion in Sequence-to-Sequence Learning

Xuebo Liu, Longyue Wang, Derek F. Wong, Liang Ding, Lidia S. Chao, Zhaopeng Tu

ICLR 2021 · Architectures and efficiency › Attention and transformers · approach: Language model pretraining and induction · 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 hard fusion of source-surface probabilityTransformer Seq2Seq · WMT16 Ro-EnBLEU: 33.8 -> 35.1Improved
Add hard fusion of source-surface probabilityTransformer Seq2Seq · WMT14 En-DeBLEU: 28.9 -> 29.5Improved
Add hard fusion of source-surface probabilityTransformer Seq2Seq · WMT14 En-FrBLEU: 43.4 -> 43.9Improved
Add soft fusion at the logit levelTransformer Seq2Seq · CNN/Daily MailROUGE-L: 37.2 -> 37.9Improved
Add soft fusion at the logit levelTransformer Seq2Seq · CONLL14F0.5: 54.3 -> 56.6Improved
Use soft rather than hard fusionSurfaceFusion · Translation benchmarksBLEUMixed
Use soft rather than hard fusionSurfaceFusion · CNN/Daily Mail and CONLL14ROUGE-L / F0.5Improved
Fuse encoder embedding features through fine-grained layer attentionVanilla Transformer · WMT16 Ro-EnBLEU: 33.80 -> 34.45Improved
Add fine-grained layer attentionVanilla Transformer · Machine translation, grammatical error correction, and summarizationTask scoresImproved
Add only the encoder embedding layer connection for the uppermost decoder layer (ablation)Fine-grained layer attention · Seq2Seq tasksTask performanceNo clear effect
Mask the encoder embedding layer (ablation)Fine-grained layer attention · Machine translation, summarization, and grammatical error correctionModel performanceHurt
Mask the encoder embedding layer (ablation)Fine-grained layer attention · Machine translation, summarization, and grammatical error correctionOutput lengthMixed

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