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

Reward Scoring on HHA benchmark

66.97Harmlessness Score (Base)

SFT+DPO

31.672440.83625059.1638Feb 16, 2025
Updated 26d ago

Evaluation Results

MethodLinks
2025.02
66.9767.8368.1668.9468.3167.97
2025.02
66.3767.7471.771.0870.1769.22
2025.02
65.1567.8671.2470.7869.9668.75
2025.02
64.9666.4669.8169.3968.5567.65
2025.02
64.8666.4170.8769.7869.0267.98
2025.02
64.2465.2369.7968.567.8466.94
2025.02
63.366.2970.0769.3268.5667.24
2025.02
62.7666.0769.1368.7867.9966.69
2025.02
61.5166.0966.567.2866.6265.35
2025.02
61.3364.7268.8768.0167.265.73
2025.02
60.7862.0261.6262.6762.161.77
2025.02
60.5863.3763.8464.964.0363.17
2025.02
58.2261.7263.5863.5662.9561.77
2025.02
56.4258.0260.3859.9159.4458.68
2025.02
55.8355.6359.0457.6157.4357.03
2025.02
55.2258.0962.3260.3860.2659
2025.02
55.2155.556.2456.7556.1755.93
2025.02
54.6158.4558.6960.1959.1157.98
2025.02
54.5161.1361.5462.2661.6459.86
2025.02
53.45757.3857.8557.4156.41
2025.02
53.1849.0755.6651.952.2152.45
2025.02
50.0345.3651.7747.7748.348.73
2025.02
47.0734.2741.8636.6137.5839.96
2025.02
42.4942.0247.9644.2144.7344.17
2025.02
39.2932.7339.0335.1635.6436.55
2025.02
38.834.3639.1336.7836.7737.27
2025.02
37.9527.728.6328.1628.1630.61
2025.02
37.0931.3834.5833.043334.02
2025.02
35.0126.1832.1128.1328.8130.36
2025.02
33.0325.4430.8626.9427.7529.06