AI Paper Atlas

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

DABS 2.0: Improved Datasets and Algorithms for Universal Self-Supervision

Alex Tamkin, Gaurab Banerjee, Mohamed Owda, Vincent Liu, Shashank Rammoorthy, Noah Goodman

NeurIPS 2022 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Data-efficient and robust self-supervision · 0 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
pretrain with MAE (50% corruption) vs no pretrainingDomain-agnostic Transformer · Satellite Imagesaverage transfer metric: 62.3 -> 86.3Improved
pretrain with ShED (15%) vs no pretrainingDomain-agnostic Transformer · Particle Physicsaverage transfer metric: 54.8 -> 68.0Improved
pretrain with ShED (15%) vs no pretrainingDomain-agnostic Transformer · Wearable Sensorsaverage transfer metric: 69.8 -> 85.2Improved
pretrain with ShED (15%) vs no pretrainingDomain-agnostic Transformer · Satellite Imagesaverage transfer metric: 62.3 -> 57.8Hurt
pretrain with ShED instead of MAEDomain-agnostic Transformer · Protein Biologyaverage transfer metric: 32.2 -> 36.8Improved
raise ShED corruption rate from 15% to 50% (ablation)ShED · Wearable Sensorsaverage transfer metric: 85.2 -> 65.2Hurt
raise ShED corruption rate from 15% to 50% (ablation)ShED · Multilingual Textaverage transfer metric: 57.5 -> 64.0Improved
raise ShED corruption rate from 15% to 85% (ablation)ShED · Natural Imagesaverage transfer metric: 16.1 -> 24.6Improved
raise MAE corruption rate from 15% to 50% (ablation)MAE · Genomicsaverage transfer metric: 26.6 -> 39.0Improved
raise MAE corruption rate from 50% to 85% (ablation)MAE · Genomicsaverage transfer metric: 39.0 -> 25.7Hurt
raise MAE corruption rate from 15% to 85% (ablation)MAE · Captioned Imagesaverage transfer metric: 51.4 -> 50.0No clear effect
use Capri (15%) vs no pretrainingDomain-agnostic Transformer · Satellite Imagesaverage transfer metric: 62.3 -> 11.1Hurt

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