AI Paper Atlas

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

Adversarial VQA: A New Benchmark for Evaluating the Robustness of VQA Models

Linjie Li, Jie Lei, Zhe Gan, Jingjing Liu

ICCV 2021 · Vision-language and multimodal › Visual question answering · approach: Universal VQA models · 66 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 R1 adversarial data to training (ablation)UNITER-B · AVQA testaccuracy (%): 18.81 -> 20.47Improved
add R1 and R2 data to training (ablation)UNITER-B · AVQA testaccuracy (%): 18.81 -> 22.62Improved
add all AVQA rounds to training (ablation)UNITER-B · AVQA testaccuracy (%): 18.81 -> 24.10Improved
add AVQA to training (ablation)UNITER-B · VQA v2 test-devaccuracy (%): 72.70 -> 72.66No clear effect
train on AVQA only instead of VQA v2+VG (ablation)UNITER-B · VQA v2 test-devaccuracy (%): 72.70 -> 59.99Hurt
add image features to the model (versus language-only) (ablation)UNITER-B · R1 testaccuracy (%): 17.82 -> 17.91Improved
train on combined data (with correctly answered examples) versus verified only (ablation)UNITER-B · R1 testaccuracy (%): 25.63 -> 26.85No clear effect
scale UNITER-B to UNITER-L (ablation)UNITER · AVQA testaccuracy (%): 24.10 -> 24.78Mixed
add random-image regularization (ablation)BUTD · AVQA valaccuracy (%): 23.91 -> 23.79Hurt
add random-image regularization (ablation)UNITER-B · AVQA valaccuracy (%): 25.04 -> 24.70Hurt
add random-image regularization (ablation)BUTD · VQA-CP v2 testaccuracy (%): 40.62 -> 43.96Improved
use grid-feature end-to-end ClipBERT instead of region-feature UNITER-B (ablation)VQA model · AVQA testaccuracy (%): 18.81 -> 21.16No clear effect

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