AI Paper Atlas

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

Multimodal Learning and Reasoning for Visual Question Answering

Ilija Ilievski, Jiashi Feng

NeurIPS 2017 · Vision-language and multimodal › Visual question answering · approach: Universal VQA models · 38 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 scene classification module (ablation)VQA baseline · VQA v2.0 valoverall accuracy: 55.13 -> 56.80Improved
add object detection module (ablation)VQA+Sc · VQA v2.0 valoverall accuracy: 56.80 -> 58.46Improved
add object classification module (ablation)VQA+Sc+oDec · VQA v2.0 valoverall accuracy: 58.46 -> 59.82Improved
add face detection module (ablation)VQA+Sc+oDec+oCls · VQA v2.0 valOther accuracy: 54.47 -> 53.79Hurt
add face detection module (ablation)VQA+Sc+oDec+oCls · VQA v2.0 valYes/No accuracy: 72.88 -> 74.21Improved
add face analysis module (ablation)VQA+...+fDec · VQA v2.0 valoverall accuracy: 60.35 -> 60.60Improved
add face analysis module (ablation)VQA+...+fDec · VQA v2.0 valOther accuracy: 53.79 -> 54.81Improved
add face analysis module (ablation)VQA+...+fDec · VQA v2.0 valYes/No accuracy: 74.21 -> 73.78Hurt
replace Hadamard product with bilinear interaction (ablation)ReasonNet · VQA v2.0 valoverall accuracy: 58.37 -> 60.60Improved
replace MCB with bilinear interaction (ablation)ReasonNet · VQA v2.0 valoverall accuracy: 58.78 -> 60.60Improved
Hadamard product vs MCB fusion (ablation)ReasonNet · VQA v2.0 valoverall accuracy: 58.37 -> 58.78No clear effect
ReasonNet vs previous best (MLB)MLB · VQA v1.0 test-standardoverall accuracy: 65.1 -> 67.9Improved

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