AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

On the Pareto Front of Multilingual Neural Machine Translation

Liang Chen, Shuming Ma, Dongdong Zhang, Furu Wei, Baobao Chang

NeurIPS 2023 · Learning theory and algorithms › Prediction and decision losses · approach: Theory of neural network sensitivity · 2 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
compute sampling ratios with Double Power Law instead of tuned temperature (T=2)MNMT base model · 2-direction En->{DE,HI}average BLEU: 23.4 -> 23.7Improved
compute sampling ratios with DPL instead of best temperature (T=5)MNMT base model · 4-directionaverage BLEU: 21.8 -> 21.9Improved
DPL instead of Gradient VaccineMNMT base model · 2-directionaverage BLEU: 23.1 -> 23.7Improved
DPL instead of PCGradMNMT base model · 4-directionaverage BLEU: 19.6 -> 21.9Improved
optimize DPL for low-resource Hindi only (r_HI=1) instead of arithmetic meanMNMT base model · 2-direction En->HIBLEU: 12.0 -> 12.6Improved
DPL instead of best temperature T=5MNMT base model · 4-directionaverage BERT-Score: 70.4 -> 70.8Improved
DPL instead of best temperature T=5MNMT base model · 4-directionaverage COMET: 49.7 -> 50.2Improved
DPL instead of best temperature T=2MNMT base model · 2-directionaverage cross-entropy: 2.00 -> 1.99No clear effect
increase temperature from T=1 to T=2 (more weight to low-resource)MNMT base model · 4-directionaverage BLEU: 19.7 -> 21.8Improved
raise temperature from T=5 to T=100MNMT base model · 2-directionaverage BLEU: 23.1 -> 18.5Hurt
model size from Base (64M) to Large (317M)Double Power Law fit · power term betaestimated beta: 1.21 -> 3.43Improved
tune gamma term of DPL to -0.25 for X->EN directions (ablation)Double Power Law fit · X->EN WMT10 low-resourcer^2: 0.86 -> 0.91Improved

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