Human Preference Evaluation
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
1.053Average Score
40
Apr 21, 2026
85.4Preference Accuracy
32
May 26, 2026
0.675Preference Accuracy
32
May 26, 2026
29.87Photo Score
24
Mar 4, 2026
33D2D Preference Count
15
Jun 24, 2026
72Wins
9
Mar 6, 2026
1,183Arena Content Score
7
Mar 16, 2026
74.2Preference Accuracy
7
Feb 26, 2026
30Win Count
5
May 6, 2026
290Wins
4
Apr 28, 2026
47.8Preference Percentage
4
Mar 4, 2026
59.3Physics Plausibility Win Rate
4
Feb 26, 2026
54.9Physics Plausibility Win Rate
4
Feb 26, 2026
0.228ImageReward
3
Jul 7, 2026
200Total Battles
3
May 13, 2026
66.05Preferred Rate
3
Apr 21, 2026
94Step Faithfulness Win Rate
3
Apr 7, 2026
23.4Win Rate
3
Feb 26, 2026
56.7Win Rate
3
Feb 26, 2026
24Category CT Score
3
Feb 26, 2026
73.5Physics Correctness
3
Feb 26, 2026
39.89Recall @1
3
Feb 26, 2026
65.14Preference Accuracy
3
Feb 26, 2026
65.5Action Following
2
Jun 9, 2026
67.5Winrate
2
Jun 4, 2026