AI Paper Atlas

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

Improving Selective Visual Question Answering by Learning From Your Peers

Corentin Dancette, Spencer Whitehead, Rishabh Maheshwary, Ramakrishna Vedantam, Stefan Scherer, Xinlei Chen, Matthieu Cord, Marcus Rohrbach

CVPR 2023 · Vision-language and multimodal › Visual question answering · approach: Universal VQA models · 10 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
Use peer-generated correctness targets to train the Selector on A+B rather than self-generated targetsCLIP-ViL · VQA v2 testcoverage at 1% risk: 6.45 -> 18.40Improved
Use peer-generated correctness targets to train the Selector on A+B rather than self-generated targetsOFA-Base · VQA v2 testcoverage at 1% risk: 26.64 -> 27.71Improved
Use peer-generated correctness targets to train the Selector on A+B rather than self-generated targetsOFA-Large · VQA v2 testcoverage at 1% risk: 31.47 -> 32.92Improved
Train a Selector instead of using the model's MaxProb confidenceOFA-Large · VQA v2 testcoverage at 1% risk: 16.31 -> 31.47Improved
Use peer-generated correctness targets to train the Selector on A+B rather than self-generated targetsCLIP-ViL · 90% VQA v2 + 10% AdVQAcoverage at 1% risk: 2.97 -> 15.22Improved
Use peer-generated correctness targets to train the Selector on A+B rather than self-generated targetsOFA-Base · 90% VQA v2 + 10% AdVQAcoverage at 1% risk: 19.72 -> 21.58Improved
Use peer-generated correctness targets to train the Selector on A+B rather than self-generated targetsOFA-Large · 90% VQA v2 + 10% AdVQAcoverage at 1% risk: 22.68 -> 25.38Improved
Add KNN OOD-detection features to the Selector (ablation)OFA-Base Selector · 90% VQA v2 + 10% AdVQAcoverage at 1% risk: 19.05 -> 19.92No clear effect
Add SSD OOD-detection features to the Selector (ablation)OFA-Base Selector · 90% VQA v2 + 10% AdVQAcoverage at 1% risk: 19.05 -> 18.99No clear effect
Add known OOD training examples from OK-VQA and VizWiz to Selector training (ablation)OFA-Base Selector · 90% VQA v2 + 10% AdVQAcoverage at 1% risk: 19.00 -> 18.48Hurt
Reduce peer splits from 10 to 2 (ablation)OFA-Base LYP · VQA v2 testcoverage at 1% risk: 27.71 -> 27.64No clear effect
Reduce Selector training data from 100% to 5% of A+B (ablation)OFA-Base LYP · VQA v2 testcoverage at 1% risk: 27.71 -> 22.62Hurt

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