programbench/20260918_mini-v2.4.5_muse-spark-1.3-max
Updated • 14
None defined yet.
Given only a compiled binary and its documentation, AI agents must architect and implement a complete codebase that reproduces the original program's behavior. ProgramBench evaluates this capability across 200 real-world open-source projects spanning Rust, Go, C, C++, Haskell, and Java.