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

Human Preference Alignment on MT-Bench

6.83SAG

RAG-Pref

3.01324.00414.9955.9859May 11, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.05
6.83--
2026.05
6.75--
2026.05
6.54--
2026.05
6.49--
2026.05
6.45--
2026.05
6.22--
2026.05
6.19--
2026.05
6.12--
2026.05
6.01--
2026.05
6--
2026.05
5.97--
2026.05
5.84--
2026.05
5.84--
2026.05
5.74--
2026.05
5.74--
2026.05
5.14--
2026.05
4.58--
2026.05
4.05--
2026.05
3.58--
2026.05
3.16--
2026.06
-5.094.41
2026.06
-64.81
2026.06
-6.565.1
2026.06
-5.985.13
2026.06
-6.925.32
2026.06
-5.564.73
2026.06
-6.885.43
2026.06
-6.895.16
2026.06
-5.564.83
2026.06
-6.795.74
2026.06
-7.436.14
2026.06
-7.056.13
2026.06
-8.226.93
2026.06
-6.245.38
2026.06
-8.096.78
2026.06
-8.186.97