Skip to content

Benchmark

UniGenBench++ (English Long) - Consistency

Scores are percentages.

Models measured
32
Scores recorded
32
Sources
1
IQ weight
1.34

Score distribution

76.6%
Mean
76.6%
Median
11%
Std dev
66.3%
10th percentile
88.9%
90th percentile
37.7%
Min
37.7%91.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.34 for image-generation models, measured over 31 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

Data credits: UniGenBench++ (CodeGoat24), MIT (licence)