AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

A Character-Level Length-Control Algorithm for Non-Autoregressive Sentence Summarization

Puyuan Liu, Xiang Zhang, Lili Mou

NeurIPS 2022 · Architectures and efficiency › Attention and transformers · approach: Language model pretraining and induction · 4 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
replace truncation with character-level DP length control (ablation)NACC · Gigaword test, supervisedROUGE-1 F1: 33.12 -> 33.66Improved
replace truncation with character-level DP length control (ablation)NACC · Gigaword test, unsupervisedROUGE-1 F1: 25.79 -> 27.45Improved
replace truncation with character-level DP length control (ablation)NACC · DUC2004ROUGE-1 Recall: 26.43 -> 28.37Improved
replace truncation with character-level DP length control (ablation)Autoregressive Transformer · Gigaword test, supervisedROUGE-1 F1: 35.12 -> 34.10Hurt
replace truncation with character-level DP length control (ablation)Autoregressive Transformer · Gigaword test, unsupervisedROUGE-1 F1: 26.31 -> 25.33Hurt
use character-constrained search instead of word-constrained search followed by truncationSchumann et al. search · Gigaword test, unsupervisedROUGE-1 F1: 24.98 -> 25.30Improved
increase the length bucket size (ablation)NACC length-control decoding · Gigaword test, unsupervisedmean ROUGEHurt
use character-length control rather than truncation (ablation)NACC · Gigaword test, unsupervised human evaluationoverall quality win rate: 18% -> 38%Improved
use character-length control rather than truncation (ablation)NACC · Gigaword test, unsupervised human evaluationcompleteness and fluency win rate: 22% -> 42%Improved

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