AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Active Learning Helps Pretrained Models Learn the Intended Task

Alex Tamkin, Dat Pham Nguyen, Salil Deshpande, Jesse Mu, Noah Goodman

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Deep active learning · 14 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
uncertainty (least confidence) sampling instead of random acquisitionpretrained BiT-M-R50x1 · Waterbirdsvalidation accuracy: +11% avg; 5x fewer labelsImproved
uncertainty sampling instead of randomRoBERTa-Large · Amazon-WILDS OODaccuracy: +1% avg; +2.5% worst 10th pctImproved
uncertainty sampling instead of randomBiT-M-R50x1 · iWildCamaccuracy: +9%; 1.8x fewer labelsImproved
uncertainty sampling instead of randomBiT-M-R50x1 · Treepersonaccuracy: +2%; 1.6x fewer labelsImproved
random initialization instead of pretraining (ablation)BiT-R50 uncertainty sampling · Waterbirds/iWildCam/Treepersonaccuracy gain from ALHurt
pretrain on ImageNet-21k instead of ImageNet-1k (ablation)BiT R50x1 + AL · iWildCamaccuracy vs randomImproved
deeper model BiT-M-R101x1 instead of R50x1 (ablation)BiT-M + AL · Waterbirds/iWildCam/TreepersonAL gainNo clear effect
stop finetuning on training-loss heuristic instead of validation plateau (ablation)uncertainty sampling BiT · Waterbirds (small experiment)accuracy: 83.64 -> 85.62Improved
apply AL on domain far from pretraining (Camelyon17, FMoW)BiT-M + uncertainty sampling · Camelyon17-WILDS / FMoW-WILDSaccuracy vs randomHurt
more mismatched examples (higher task ambiguity) (ablation)BiT-M + uncertainty sampling · Waterbirds variantsAL gain over randomImproved

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