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

Preference Alignment on PRISM

74.5Win-Rate (DPO)

CUMA

54.42859.63964.8570.061Jan 8, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.01
74.576.8
2026.01
73.275.5
2026.01
71.273.5
2026.01
68.867.5
2026.01
68.570.4
2026.01
67.569.2
2026.01
65.568.2
2026.01
64.265.8
2026.01
63.565.2
2026.01
62.864.1
2026.01
61.563.2
2026.01
61.562.2
2026.01
59.561.2
2026.01
58.859.5
2026.01
58.459.5
2026.01
57.558.8
2026.01
56.559.2
2026.01
56.257
2026.01
55.556.8
2026.01
55.255.8