Add Python benchmark data (120-task hard subset)
#2
by boyuzhuhf - opened
No description provided.
boyuzhuhf changed pull request title from Write permission probe to Add Python benchmark data (120-task hard subset)
Python CyberGym task archives for the 120 unsolved (budget_exhausted) cases of the level1 full-870 run.
Layout (tar.gz only, matching the JavaScript dataset format):
python/benchmarks/tasks/<CVE>/images/<CVE>-{vul,fix}.tar.gzpython/benchmarks/tasks/<CVE>/levels/level{0,1,2,3}/repo-vul.tar.gzpython/benchmarks/tasks/<CVE>/levels/level3/repo-fix.tar.gzpython/benchmarks/tasks/<CVE>/source/repo-{vul,fix}.tar.gz
120 tasks, 1080 archives (~40.2 GB). One note: CVE-2022-0273's source/repo-vul.tar.gz was absent in the build tree; the byte-identical levels/level1/repo-vul.tar.gz (sha256 verified against metadata) is included in its place.
guan-wang changed pull request status to merged