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

Reward Model Transfer on HelpSteer3 (H3) v1 (test)

8.96AOG

ANCHOR

-6.6088-2.56691.4755.5169May 25, 2026
Updated 8d ago

Evaluation Results

MethodLinks
2026.05
8.96--8.96
2026.05
6.78--6.78
2026.05
6.01--6.01
2026.05
5.79--5.79
2026.05
5.57--5.57
2026.05
4.59--4.59
2026.05
3.83--3.83
2026.05
3.61--3.2
2026.05
3.39--3.39
2026.05
3.28--3.28
2026.05
2.08--2.08
2026.05
1.42--1.24
2026.05
0.55---2.87
2026.05
-1.85---4.23
2026.05
-5.68---12.33
2026.05
-6.01---19.05
2026.05
-32.231.71-
2026.05
--39.9-2.12-
2026.05
--295.1-15.67-
2026.05
-67.23.57-
2026.05
--87.7-3.06-
2026.05
--65.9-2.3-
2026.05
--332.1-11.59-
2026.05
-9.50.33-