AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

The Pursuit of Human Labeling: A New Perspective on Unsupervised Learning

Artyom Gadetsky, Maria Brbic

NeurIPS 2023 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Self-supervision in supervised pipelines · 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
use transductive instead of inductive setting (DINO)HUME · STL-10ACC: 90.8 -> 93.2Improved
use DINOv2 instead of BiT as the regularizing representation (inductive) (ablation)HUME · CIFAR-100-20ACC: 47.8 -> 55.5Improved
use CLIP instead of BiT as the second representation (inductive) (ablation)HUME · STL-10ACC: 87.5 -> 90.2Improved
compare unsupervised HUME with supervised linear probeHUME · CIFAR-100-20ACC: 72.5 -> 55.5Hurt
compare unsupervised HUME with supervised linear probeHUME · STL-10ACC: 88.9 -> 93.2Improved
replace SPICE with HUMEdeep clustering · STL-10ACC: 86.2 -> 90.8Improved
replace SPICE with HUMEdeep clustering · CIFAR-100-20ACC: 46.8 -> 55.5Improved
replace best baseline with HUMEdeep clustering · ImageNet-1000ACC: 41.1 -> 51.1Improved
use HUME reliable samples instead of random ground truth labelsFreeMatch · CIFAR-10, 1 label/classaccuracy: 83.3 -> 91.2Improved
use HUME reliable samples instead of random ground truth labelsFreeMatch · STL-10, 100 labels/classaccuracy: 94.0 -> 91.3Hurt
use HUME reliable samples instead of ground truth labelsFreeMatch · CIFAR-10, 400 labels/classaccuracy: 95.7 -> 94.3Hurt
use MOCO instead of SimCLR representation (ablation)HUME · STL-10ACC: 86.9 -> 90.8Improved

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