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

Accuracy (%) on Reasoning on HLE

40.8Accuracy (HLE Reasoning)

DeepSeek-V3.2

-0.269610.392721.05531.7173Mar 25, 2026Mar 29, 2026Apr 3, 2026Apr 8, 2026Apr 12, 2026Apr 17, 2026Apr 22, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
40.8
2026.04
35.2
2026.04
32.9
2026.04
31.8
2026.04
30.4
2026.04
29.8
2026.04
28.7
2026.04
26.9
2026.04
25.3
2026.04
24.5
2026.04
22.82
2026.04
22
2026.04
21.7
2026.04
21.5
2026.04
20.3
2026.04
19.1
2026.04
19
2026.04
18.26
2026.04
18
2026.04
17.3
2026.04
14.9
2026.03
7.71
2026.03
6.51
2026.03
5.63
2026.03
5.45
2026.04
5.3
2026.03
5.15
2026.03
4.96
2026.04
4.9
2026.04
4.8
2026.04
4.8
2026.04
4.8
2026.03
4.66
2026.04
4.6
2026.04
4.6
2026.04
4.6
2026.04
4.6
2026.03
4.45
2026.04
4.4
2026.04
4.4
2026.03
4.34
2026.04
4.2
2026.03
4.02
2026.03
3.95
2026.04
3.9
2026.04
3.7
2026.04
3.7
2026.04
3.7
2026.04
3.5
2026.03
3.37
2026.03
3.29
2026.03
3.12
2026.04
3
2026.04
2.8
2026.03
2.73
2026.03
2.56
2026.04
2.5
2026.04
2.5
2026.04
2.5
2026.03
2.34
2026.04
1.9
2026.04
1.9
2026.03
1.31