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

Preference Alignment on AlignX (Arbitrary)

74.6Accuracy

Raw-Aligned

49.01655.65862.368.942Oct 17, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.10
74.61,402.86
2025.10
74.41259.92
2025.10
71.12151.65
2025.10
70.56257.52
2025.10
69.56288.75
2025.10
69.4591.47
2025.10
56.5591.47
2025.10
54.63151.65
2025.10
52.69257.52
2025.10
52.1288.75
2025.10
51.97259.92
2025.10
51.891,402.86
2025.10
50-
2025.10
50-