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

Preference Alignment on HH-RLHF (test)

87.4Win Rate

CW-IPO

53.714462.459771.20579.9503Mar 5, 2026
Updated 1d ago

Evaluation Results

MethodLinks
2026.03
87.4--
2026.03
86.8--
2026.03
86.64--
2026.03
86.2--
2026.03
84.16--
2026.03
83.4--
2026.03
83.01--
2026.03
82.26--
2026.03
82.2--
2026.03
82.07--
2026.03
81.88--
2026.03
81.5--
2026.03
81.4--
2026.03
81.2--
2026.03
81--
2026.03
80.6--
2026.03
79.77--
2026.03
78.8--
2026.03
64.3--
2026.03
63.5--
2026.03
63--
2026.03
62.95--
2026.03
62.8--
2026.03
61.79--
2026.03
61.3--
2026.03
60.79--
2026.03
58.87--
2026.03
58.2--
2026.03
57.63--
2026.03
57.6--
2026.03
56.9--
2026.03
56.88--
2026.03
56.7--
2026.03
55.92--
2026.03
55.9--
2026.03
55.01--
2026.05
-48.1277.34
2026.05
-48.8575.12
2026.05
-62.2484.56
2026.05
-66.4589.21
2026.05
-43.0569.22
2026.05
-44.2171.45
2026.05
-54.1878.12
2026.05
-58.3284.05
2026.05
-41.4867.15
2026.05
-42.3369.42
2026.05
-53.0577.34
2026.05
-55.1882.31
2026.05
-52.3483.12
2026.05
-51.6582.45
2026.05
-66.8289.15
2026.05
-73.2196.34
2026.05
-44.1578.45
2026.05
-46.2178.12
2026.05
-56.4184.34
2026.05
-64.4895.12
2026.05
-41.2275.33
2026.05
-43.4476.05
2026.05
-53.2580.42
2026.05
-60.1592.45