AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Are Disentangled Representations Helpful for Abstract Visual Reasoning?

Sjoerd van Steenkiste, Francesco Locatello, Jürgen Schmidhuber, Olivier Bachem

NeurIPS 2019 · Vision-language and multimodal › Human and embodied modeling · approach: Abstract visual reasoning benchmarks · 116 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
Use more disentangled pretrained representations instead of less disentangled representationsWReN abstract reasoning models · dSprites and 3dshapes abstract visual reasoningdownstream accuracy in the few-sample regimeImproved
Use representations with higher FactorVAE scoresWReN abstract reasoning models · dSprites and 3dshapes abstract visual reasoning, up to 20K stepsdownstream accuracyImproved
Use representations with higher BetaVAE scoresWReN abstract reasoning models · dSprites and 3dshapes abstract visual reasoning, up to 20K stepsdownstream accuracyImproved
Train with more labeled samples, from the few-sample to the many-sample regimeWReN using disentangled representations · dSprites and 3dshapes abstract visual reasoningcorrelation between disentanglement and downstream accuracyHurt
Use higher reconstruction quality instead of higher disentanglementWReN abstract reasoning models · dSprites and 3dshapes abstract visual reasoning, few-sample regimecorrelation with downstream accuracyHurt
Use higher reconstruction qualityWReN abstract reasoning models · dSprites and 3dshapes abstract visual reasoning, many-sample regimecorrelation with downstream accuracyImproved
Use representations ranked by DCI Disentanglement rather than by FactorVAE or BetaVAE scoreWReN abstract reasoning models · dSprites and 3dshapes abstract visual reasoning, few-sample regimecorrelation with downstream accuracyHurt
Use representations ranked by MIG or SAP rather than by FactorVAE or BetaVAE scoreWReN abstract reasoning models · dSprites and 3dshapes abstract visual reasoning, few-sample regimecorrelation with downstream accuracyHurt

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