AI Paper Atlas

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

A Benchmark for Compositional Visual Reasoning

Aimen Zerroug, Mohit Vaishnav, Julien Colin, Sebastian Musslick, Thomas Serre

NeurIPS 2022 · Vision-language and multimodal › Human and embodied modeling · approach: Abstract visual reasoning benchmarks · 7 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
add SSL pretraining, joint trainingResNet-50 · CVR 1000 samplesaccuracy: 59.2 -> 79.2Improved
add SSL pretraining, individual trainingViT-small · CVR 1000 samplesaccuracy: 35.1 -> 65.5Improved
add SSL pretraining, joint trainingViT-small · CVR 1000 samplesaccuracy: 31.4 -> 60.7Improved
train jointly on all rules instead of individuallyResNet-50 · CVR 1000 samplesaccuracy: 44.8 -> 59.2Improved
train jointly on all rules instead of individuallyViT-small · CVR 1000 samplesaccuracy: 35.1 -> 31.4Hurt
add ResNet-18 backbone to SCLSCL · CVR 1000 samples individualaccuracy: 33.4 -> 48.3Improved
joint training with SSL pretraining (SES)ResNet-50 vs ViT-small · CVRSES: 57.0 vs 44.7Improved
increase training data from 1000 to 10k with SSL jointResNet-50 · CVRaccuracy: 79.2 -> 93.1Improved
pretrain on elementary rules then finetune on compositions (curriculum)baselines · CVR compositionsaccuracyImproved
test composition-trained models on elementary rules (reverse curriculum)baselines · CVR elementary rulesaccuracyHurt
compare models with humans at 20 samplesResNet-50 SSL vs humans · CVR 45 tasksaccuracy: 45.7 vs 78.7Hurt
increase training samples from 20 to 1000ResNet-50 SSL · CVR 45 tasksaccuracy: 45.7 -> 78.3Improved

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