ProgramBench: Can Language Models Rebuild Programs From Scratch?

TL;DR

ProgramBench is a new benchmark that evaluates language models' ability to develop complete software projects from scratch based on minimal input, revealing current models' limitations in holistic software engineering tasks.

Contribution

The paper introduces ProgramBench, a comprehensive benchmark for assessing language models' capacity to architect and implement entire software systems from scratch.

Findings

No model fully solves any task in the benchmark.

The best model passes 95% of tests on only 3% of tasks.

Models tend to produce monolithic, single-file implementations.

Abstract

Turning ideas into full software projects from scratch has become a popular use case for language models. Agents are being deployed to seed, maintain, and grow codebases over extended periods with minimal human oversight. Such settings require models to make high-level software architecture decisions. However, existing benchmarks measure focused, limited tasks such as fixing a single bug or developing a single, specified feature. We therefore introduce ProgramBench to measure the ability of software engineering agents to develop software holisitically. In ProgramBench, given only a program and its documentation, agents must architect and implement a codebase that matches the reference executable's behavior. End-to-end behavioral tests are generated via agent-driven fuzzing, enabling evaluation without prescribing implementation structure. Our 200 tasks range from compact CLI tools to widely used software such as FFmpeg, SQLite, and the PHP interpreter. We evaluate 9 LMs and find that none fully resolve any task, with the best model passing 95\% of tests on only 3\% of tasks. Models favor monolithic, single-file implementations that diverge sharply from human-written code.