AI Paper Atlas

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

A Case Study of the Shortcut Effects in Visual Commonsense Reasoning

Keren Ye, Adriana Kovashka

AAAI 2021 · Vision-language and multimodal › Visual question answering · approach: Visual reasoning diagnostics · 32 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
Replace person tags and pronouns in answer options with semantically equivalent referencesTAB-VCR · VCR Q→A, RULE-SINGULARaccuracy: 70.5 -> 65.4Hurt
Replace person tags and pronouns in answer options with semantically equivalent referencesTAB-VCR · VCR QA→R, RULE-SINGULARaccuracy: 72.4 -> 66.3Hurt
Mask the most influential answer tokenVCR models · VCR Q→A, ADVTOP-1accuracy: STD VAL 63.8–69.9; ADVTOP-1 49.8–54.9Hurt
Train B2T2 with curriculum masking, initial ratio 0.30 and decay rate 5E-5 (ablation)B2T2 · VCR Q→A, STD VALaccuracy: 68.5 -> 69.9Improved
Train B2T2 with curriculum masking, initial ratio 0.30 and decay rate 5E-5 (ablation)B2T2 · VCR Q→A, ADVTOP-1accuracy: 37.0 -> 54.5Improved
Train B2T2 with curriculum masking, initial ratio 0.30 and decay rate 5E-5 (ablation)B2T2 · VCR QA→R, STD VALaccuracy: 68.5 -> 70.6Improved
Use static random answer-token masking at probability 0.30 (ablation)B2T2 · VCR Q→A, STD VALaccuracy: 68.5 -> 64.1Hurt
Use static random answer-token masking at probability 0.30 (ablation)B2T2 · VCR Q→A, ADVTOP-1accuracy: 37.0 -> 47.5Improved
Combine static answer-token masking at probability 0.10 with masked language modeling (ablation)B2T2 · VCR Q→A, STD VALaccuracy: 68.5 -> 69.1Improved
Augment training with rule-based modifications to person references (ablation)B2T2 · VCR Q→A, RULE-SINGULARaccuracy: 63.3 -> 78.8Improved
Augment training with rule-based modifications to person references (ablation)B2T2 · VCR Q→A, ADVTOP-1accuracy: 37.0 -> 31.6Hurt
Augment training with adversarially masked answer tokens (ablation)B2T2 · VCR Q→A, ADVTOP-1accuracy: 37.0 -> 81.4Improved

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