AI Paper Atlas

Research map / Vision-language and multimodal / Visual question answering

Transformer Module Networks for Systematic Generalization in Visual Question Answering

Moyuru Yamada, Vanessa D’Amario, Kentaro Takemoto, Xavier Boix, Tomotake Sasaki

TPAMI 2024 · Vision-language and multimodal › Visual question answering · approach: Attention and prior-robust VQA · 12 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
modularize Transformer with question-specific module composition (TMN-Tree vs Transformer)Transformer · CLOSUREaccuracy: 57.4 -> 95.4Improved
use TMN-Stack instead of TransformerTransformer · CLOSUREaccuracy: 57.4 -> 90.9Improved
use TMN-Tree vs Transformer on GQA-SGLTransformer · GQA-SGLaccuracy: 47.7 -> 53.7Improved
use TMN on novel visual attribute combinationsTransformer · CLEVR-CoGenT condition Baccuracy: 78.9 -> 80.6No clear effect
add program-related (PR) tokens to TransformerTransformer · CLOSUREaccuracy: 57.4 -> 64.5Improved
add specialization of modules (SoM) to variable-layer split-token Transformer (ablation)Transformer w/ PR + VL + ST · CLOSUREaccuracy: 43.6 -> 90.9Improved
add variable number of layers (VL) without specialization (ablation)Transformer w/ PR · CLOSUREaccuracy: 64.5 -> 60.2Hurt
add split tokens (ST) without specialization (ablation)Transformer w/ PR + VL · CLOSUREaccuracy: 60.2 -> 43.6Hurt
assign modules to individual sub-tasks vs order-based modules (ablation)TMN-Stack · CLOSUREaccuracy: 68.4 -> 90.9Improved
group semantically similar sub-tasks into shared modules (ablation)TMN-Stack · CLOSUREaccuracy: 90.9 -> 93.7Improved
use object detector pre-trained on Visual Genome regional features (ablation)TMN-Tree · CLOSUREaccuracy: 95.4 -> 96.4Improved
use object detector pre-trained on Visual Genome (ablation)Transformer · CLEVR-CoGenT condition Baccuracy: 78.9 -> 87.5Improved

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