Commit0: Library Generation from Scratch
This addresses the problem of benchmarking generative AI systems for complex, multi-stage software development tasks, though it is incremental in extending existing code generation benchmarks.
The authors introduced Commit0, a benchmark for evaluating AI agents' ability to generate complete software libraries from scratch based on API specifications and unit tests. Their experiments showed that current agents can pass some tests but cannot fully reproduce libraries, though interactive feedback significantly improves test-passing rates.
With the goal of benchmarking generative systems beyond expert software development ability, we introduce Commit0, a benchmark that challenges AI agents to write libraries from scratch. Agents are provided with a specification document outlining the library's API as well as a suite of interactive unit tests, with the goal of producing an implementation of this API accordingly. The implementation is validated through running these unit tests. As a benchmark, Commit0 is designed to move beyond static one-shot code generation towards agents that must process long-form natural language specifications, adapt to multi-stage feedback, and generate code with complex dependencies. Commit0 also offers an interactive environment where models receive static analysis and execution feedback on the code they generate. Our experiments demonstrate that while current agents can pass some unit tests, none can yet fully reproduce full libraries. Results also show that interactive feedback is quite useful for models to generate code that passes more unit tests, validating the benchmarks that facilitate its use.