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

Open-book generation under knowledge conflict on ConFiQA 1,500 subset

81.07Ps Score

ContextDPO

9.72628.24846.7765.292Jan 7, 2026
Updated 3mo ago

Evaluation Results

MethodLinks
2026.01
81.078.879.86
2026.01
77.537.68.93
2026.01
77.410.1311.58
2026.01
75.1311.5313.31
2026.01
72.9313.3315.46
2026.01
7113.9316.41
2026.01
70.879.2711.56
2026.01
70.415.417.95
2026.01
69.6714.2717
2026.01
67.816.7319.79
2026.01
66.2715.5318.99
2026.01
62.1321.8726.03
2026.01
59.1317.3322.67
2026.01
58.9322.8727.95
2026.01
58.418.624.16
2026.01
56.9317.223.2
2026.01
53.3310.216.05
2026.01
50.615.5323.49
2026.01
48.610.2717.44
2026.01
46.2717.227.1
2026.01
42.4716.427.86
2026.01
37.0716.6731.02
2026.01
36.617.5332.39
2026.01
35.4732.647.89
2026.01
35.2732.3347.83
2026.01
35.1319.3335.5
2026.01
32.8718.4735.97
2026.01
3018.5338.19
2026.01
25.9318.441.5
2026.01
19.6719.3349.57
2026.01
14.819.857.23
2026.01
12.4717.257.98