AI Paper Atlas

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

Break It Down: Evidence for Structural Compositionality in Neural Networks

Michael A. Lepori, Thomas Serre, Ellie Pavlick

NeurIPS 2023 · Vision-language and multimodal › Human and embodied modeling · approach: Abstract visual reasoning benchmarks · 6 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
Discover sparse subnetworks for individual task subroutines and ablate each discovered subnetwork (ablation)Trained ResNet50 and BERT-Small models · Compositional vision and language tasksdifference in target-versus-other subroutine accuracyImproved
Initialize BERT-Small with pretrained language-model weights rather than training from scratchBERT-Small · Language tasksmodularity of discovered subnetworksImproved
Initialize ResNet50 with SimCLR-pretrained weights rather than random initializationResNet50 · Vision tasksstructural compositionalityNo clear effect
Ablate the Contact subnetwork (ablation)ResNet50 trained on Number-Contact · Number-Contact, Contact subroutinetarget-versus-other subroutine accuracy differenceImproved
Ablate the Number subnetwork (ablation)ResNet50 trained on Number-Contact · Number-Contact, Number subroutinetarget-versus-other subroutine accuracy differenceNo clear effect
Use a ViT backbone for the vision compositional tasksViT · Inside-Contact, Number-Contact, and Inside-NumberaccuracyHurt
Use sparse subnetworks to test whether each known subroutine is independently representedNeural networks trained on compositional vision and language tasks · Compositional vision and language tasksevidence of structural compositionalityImproved

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