Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

VCR

Benchmarks

Task NameDataset NameSOTA ResultTrend
Visual Commonsense ReasoningVCR (val)
Accuracy82.57
63
Visual Commonsense ReasoningVCR-EN-Easy
EM95.7
27
Visual Commonsense Reasoning (Q→A)VCR Shortcut Mitigated Evaluation SM (test)
Q->A Accuracy (100%)76.32
13
Visual Commonsense Reasoning (Q→A)VCR Standard Evaluation (test)
Q->A Accuracy (3%, 1000 SP/C)61.49
13
Visual Commonsense Reasoning Q→AVCR (val)
Accuracy79.4
8
Holistic Reasoning (Q -> AR)VCR (val)
Accuracy55
8
Rationale Selection (QA -> R)VCR (val)
Accuracy76
8
Visual Question Answering (Q -> A)VCR (val)
Accuracy73.7
8
Visual Commonsense ReasoningVCR 1.0 (test)
Q->A Accuracy59
7
Visual Commonsense ReasoningVCR QR -> A (val)
Accuracy94.7
7
Visual Commonsense ReasoningVCR 1.0 (val)
Q->A Accuracy80.6
7
Visual Commonsense ReasoningVCR e-ViL (test)
Meteor Score26.4
6
Visual Question Answering (Q -> A)VCR (test)
Accuracy74
6
Visual Commonsense ReasoningVCR OC 1
Accuracy36.93
4
Visual Commonsense ReasoningVCR1 MCI
Accuracy60.01
4
Visual Commonsense ReasoningVCR
Q->A Accuracy79.8
4
Explanation GenerationVCR
Preference Rate (Ours)26.4
1
Showing 17 of 17 rows