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

Knowledge Retention on MMLU (full)

88.3MMLU Accuracy

Original

24.44441.02257.674.178Oct 22, 2025Nov 21, 2025Dec 22, 2025Jan 21, 2026Feb 21, 2026Mar 23, 2026Apr 23, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
88.3
2026.04
86.3
2026.04
80.9
2026.04
80.7
2026.04
78.2
2026.04
77.4
2026.04
77.2
2026.04
76.4
2026.04
76.2
2026.04
76.2
2026.04
76
2026.04
75.3
2026.04
74.9
2026.04
74.2
2026.04
73.8
2026.04
73.7
2026.04
73.6
2026.04
73.6
2026.04
73.4
2026.04
72.8
2026.04
72.5
2026.04
71.9
2026.04
71.9
2026.04
71.3
2026.04
70.8
2026.04
70.6
2026.04
69.2
2026.04
68.8
2026.04
68
2026.04
66.7
2026.04
65.6
2026.04
65.4
2026.04
64.5
2026.04
64.1
2026.04
63.3
2026.04
62.7
2026.04
62.6
2026.04
62.5
2026.04
61.9
2026.04
61
2025.10
58
2025.10
55
2025.10
54
2025.10
52
2025.10
48
2026.03
47.8
2026.03
46.3
2026.03
46
2026.03
44.6
2026.03
44.6
2026.03
44.4
2025.10
44
2025.10
44
2026.03
43.8
2026.03
43.6
2025.10
43
2026.03
42.7
2026.03
41.1
2026.03
38.9
2026.03
26.9