From: Neutron: an attention-based neural decompiler
Training data set
Arithmetic
Logic
Comparison
Function call
M(R)
1.89%
2.35%
0.74%
10.96%
M(R+E)
82.3%
69.8%
60.77%
83.16%
M(R+E+M)
85.2%
72.06%
63.4%
87.41%