Benchmark
SciCode
Test code for scientific calculations, numerical methods, and simulations.
Published leaderboard
Explore scores and evaluation settings on the publisher’s leaderboard.
Open leaderboardCoverage
80 research coding problems with 338 subproblems; the official test set and main-problem versus subproblem metrics must be distinguished.
Methodology
Scientist-written tests check generated code. Evaluation settings vary in whether scientific background and earlier subproblem solutions are supplied.