Mahal Systems

Overview › Benchmark catalog

HumanEvalFix (HumanEvalPack)

A tiny bug-fix template: 164 HumanEval problems times six languages with planted bugs; useful only as a quick multi-language smoke test for a repair agent.

What it asks the agent to do

  • fix a bug in a short function given failing tests, in six languages
  • explain a function and re-synthesize it from the explanation

What to borrow from it

case template

Caveats

Bugs are hand-planted in toy functions, no repository context, and HumanEval derivatives are saturated and memorized.

Details

Workloadscoding
Interactionsingle turn
Environmentstatic
End state checkedNo
Toolsnone
Gradingunit tests
Size984
Datasynthetic
Personal datanone
LicenseMIT
Harnessown repo
Statusfrozen

Links