Preference Alignment
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
89.5Win Rate
45
Jun 26, 2026
68.8Win Rate
36
Mar 6, 2026
87.4Win Rate
36
Jun 1, 2026
83.2Win Rate
32
Jul 7, 2026
7.6Reward
30
Jun 4, 2026
5.9Reward (R)
30
Jun 4, 2026
80Win Rate (W)
24
Jun 4, 2026
83Win Rate
24
Jun 4, 2026
62.46Accuracy (%)
20
Mar 16, 2026
63.71Accuracy
20
Mar 16, 2026
74.5Win-Rate (DPO)
20
Feb 26, 2026
81.05Win Rate
18
Mar 6, 2026
610.7Delta Reward
15
Jul 3, 2026
196Wins (Count)
14
May 11, 2026
58.76Accuracy
14
Apr 28, 2026
59.78Accuracy
14
Apr 28, 2026
92.51Accuracy
14
Apr 28, 2026
74.6Accuracy
14
Apr 28, 2026
1.39Helpfulness (RM)
13
Feb 26, 2026
5.83LLM-as-a-Judge Helpful Score
12
Apr 8, 2026
52Win Rate
12
Mar 16, 2026
78.87Accuracy
12
Feb 26, 2026
78.8Accuracy
12
Feb 26, 2026
74.18Accuracy
11
Mar 23, 2026
94.32Accuracy
11
Mar 23, 2026