BuildBench: Benchmarking LLM Agents on Compiling Real-World Open-Source Software
A new benchmark evaluates LLM agents on compiling real-world open-source software, highlighting progress in autonomous coding capabilities.
A new benchmark evaluates LLM agents on compiling real-world open-source software, highlighting progress in autonomous coding capabilities.
Auto-scored: The benchmark demonstrates incremental capability improvement in LLM agents for software compilation, with minor potential for misuse.