AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

DoReMi: Optimizing Data Mixtures Speeds Up Language Model Pretraining

Sang Michael Xie, Hieu Pham, Xuanyi Dong, Nan Du, Hanxiao Liu, Yifeng Lu, Percy Liang, Quoc V Le, Tengyu Ma, Adams Wei Yu

NeurIPS 2023 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Mixture and distillation for domain shift · 14 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
reweight Pile domains with DoReMi (280M proxy) instead of default weights8B LM · The Pile, avg one-shot accuracyavg exact match: 20.03 -> 26.56Improved
reweight Pile domains with DoReMi (280M proxy)8B LM · The Pile valavg log-ppl: 1.64 -> 1.40Improved
increase proxy model size from 70M to 280M (ablation)DoReMi with 8B main model · The Pile, avg one-shot accuracyavg exact match: 21.11 -> 26.56Improved
increase proxy model size from 280M to 1B (ablation)DoReMi with 8B main model · The Pile, avg one-shot accuracyavg exact match: 26.56 -> 22.35Hurt
use excess loss instead of only proxy loss (hardest domains) (ablation)DoReMi 280M · The Pile valavg log-ppl: 2.62 -> 2.13Improved
use excess loss instead of only negative reference loss (easiest domains) (ablation)DoReMi 280M · The Pile valavg log-ppl: 4.18 -> 2.13Improved
match proxy and main size, DoReMi at 280M280M LM · The Pile valavg log-ppl: 2.32 -> 2.13Improved
match proxy and main size, DoReMi at 510M510M LM · The Pile valworst-case log-ppl: 2.14 -> 2.14No clear effect
use trained 1B proxy model directly vs main model with DoReMi weights (ablation)1B LM · The Pile valavg log-ppl: 2.02 vs 1.83Hurt
iterate DoReMi to round 2 on GLaM dataset (ablation)8B LM · GLaM dataset downstreamavg one-shot accuracyImproved

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