Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

SIQA

Benchmarks

Task NameDataset NameSOTA ResultTrend
Commonsense ReasoningSIQA
Accuracy89.85
168
Social Interaction Question AnsweringSIQA
Accuracy86.9
157
Social Commonsense ReasoningSIQA
Accuracy86.9
112
ReasoningSIQA
Accuracy83.2
44
Social Commonsense ReasoningSIQA (test)
Accuracy83.3
20
Social Commonsense Question AnsweringSIQA
Accuracy80.04
14
Scientific Image Quality Assessment UnderstandingSIQA-U
Scientific Completeness60.5
14
Zero-shot Common Sense ReasoningSIQA
Accuracy (Zero-shot)41.91
12
ReasoningSIQA (leave-one-out setup)
Average Accuracy82.4
12
Scientific Image Quality AssessmentSIQA-S 1.0 (test)
Perception SRCC0.857
12
ReasoningSIQA
Accuracy Improvement2.12
12
Social Interaction Question AnsweringSIQA
Normalized PLL Score50.4
10
ReasoningSIQA (val)
Accuracy35.47
9
Scaling Law FittingSiQA
Score at Scaling Factor 1e-50.987
7
Reward PredictionSIQA (out-of-domain)
Accuracy76.89
6
Commonsense ReasoningSIQA (test)
Accuracy40.28
6
Social ReasoningSIQA
Performance (%)15.2
6
Showing 17 of 17 rows