AI Paper Atlas

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

It's Not About the Journey; It's About the Destination: Following Soft Paths Under Question-Guidance for Visual Reasoning

Monica Haurilet, Alina Roitberg, Rainer Stiefelhagen

CVPR 2019 · Vision-language and multimodal › Visual question answering · approach: Visual reasoning diagnostics · 19 citations (OpenAlex)

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 question-guided graph traversal and predict from destination nodesVisual reasoning on COG · COG test, Allaccuracy: 93.7 -> 97.2Improved
Use question-guided graph traversal and predict from destination nodesVisual reasoning on AI2D · AI2Daccuracy: 41.55 -> 43.45Improved
Use question-guided graph traversal and predict from destination nodesVisual reasoning on CLEVR · CLEVR test, Allaccuracy: 98.9 -> 97.5Hurt
Use question-guided graph traversal and predict from destination nodesVisual reasoning on CLEVR · CLEVR test, Compare Numbersaccuracy: 96.8 -> 99.6Improved
Use question-guided graph traversal and predict from destination nodesVisual reasoning on CLEVR · CLEVR test, Compare Attributesaccuracy: 99.1 -> 99.8Improved
Use question-guided graph traversal and predict from destination nodesVisual reasoning on CLEVR · CLEVR test, Countaccuracy: 97.1 -> 91.3Hurt
Use question-guided graph traversal and predict from destination nodesVisual reasoning on CLEVR · CLEVR test, Existaccuracy: 99.3 -> 98.6Hurt
Use question-guided graph traversal and predict from destination nodesVisual reasoning on CLEVR · CLEVR test, Query Attributesaccuracy: 99.1 -> 99.5Improved
Limit the maximal soft-path length to 1 instead of 3 (ablation)Ours on CLEVR · CLEVR validation, Query Attributesaccuracy: 99.8% -> 53.1%Hurt
Limit the maximal soft-path length to 2 instead of 3 (ablation)Ours on CLEVR · CLEVR validation, Query Attributesaccuracy: 99.8% -> 98.1%Hurt
Train on counting questions, then evaluate on unseen query-attribute questions without further trainingOurs on CLEVR · CLEVR knowledge transfer, Query Attributesaccuracy: 30.6% -> 56.4%Improved
Train on one question type and evaluate on a different type without further trainingOurs on CLEVR · CLEVR knowledge transfer, unseen tasksaccuracyImproved

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