Software Engineering
Decompile-Bench: released IDA and Ghidra outputs
Released decompiler outputs for individual C/C++ functions from HumanEval, MBPP and GitHub2025. Derive the authors' edit-similarity metric deterministically from the original outputs and reference source. This metric measures textual similarity, not executable correctness.
72,059items
2subjects
CC0-1.0 for the released evaluation dataset; MIT for the metric implementation. Original C/C++ projects retain their source licenses.license
software_engineeringdomain
textmodality
item-level responses released
Saturation status: No
Response matrix
Loading response matrix…
lowhighUnobserved
Scale: Score in [0, 1]