Benchmark
SlopCodeBench
Test whether coding agents preserve working behavior and code quality as requirements evolve.
Published leaderboard
Explore scores and evaluation settings on the publisher’s leaderboard.
Open leaderboardCoverage
36 programming problems with 196 successive checkpoints in benchmark v1.0.
Methodology
Ask agents to extend their earlier implementations. Measure correctness across checkpoints alongside verbosity and structural erosion, preserving the exact campaign and scoring policy.