AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Order Matters in the Presence of Dataset Imbalance for Multilingual Learning

Dami Choi, Derrick Xin, Hamid Dadkhahi, Justin Gilmer, Ankush Garg, Orhan Firat, Chih-Kuan Yeh, Andrew M. Dai, Behrooz Ghorbani

NeurIPS 2023 · Architectures and efficiency › Pruning and efficient tuning · approach: Parameter-efficient transfer analysis

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
pre-train on En->Fr then jointly fine-tune on En->Ro + En->Frscalarization · En->{Ro,Fr} NMTlow-resource (En->Ro) loss trade-off frontImproved
pre-train on high-resource pair then jointly fine-tune on En->Hi + En->Frscalarization · En->{Hi,Fr} NMTEn->Hi loss trade-off frontImproved
pre-train on En->Fr before training En->Zh + En->Fr (both high-resource)scalarization · En->{Zh,Fr} NMTloss trade-off frontNo clear effect
add optimizer and LR-schedule reset to scalarization (restart baseline) (ablation)scalarization · En->{Ro,Fr} NMTEn->Ro lossNo clear effect
fine-tune only on low-resource task after pre-training (classic transfer) instead of joint fine-tuning (ablation)pre-training joint fine-tuning · En->{Ro,Fr} NMTlossHurt
increase regularization strength (dropout) in scalarization instead of pre-training (ablation)scalarization · En->{Ro,Fr} NMTEn->Ro lossNo clear effect
pre-train English then jointly fine-tune 5 languagesmT5-XXL temperature sampling (tau=3.33) · mC4 Gu, Gd, Swlanguage modeling lossImproved
pre-train English then jointly fine-tune 5 languagesmT5-XXL temperature sampling (tau=3.33) · mC4 Englishlanguage modeling lossImproved
add learning-rate schedule restart to static sampling (ablation)mT5-XXL static sampling · mC4language modeling lossImproved
lengthen pre-training under fixed 500k total steps (ablation)mT5-XL · mC4 Gu, Gd, Swlanguage modeling lossImproved
lengthen pre-training under fixed 500k total steps (ablation)mT5-XL · mC4 English and Hindilanguage modeling lossHurt
lengthen pre-training on En->Fr with unfixed budget (ablation)NMT pre-training joint fine-tuning · En->{Ro,Fr} NMToverall trade-off frontImproved

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