BUILD-BENCH

Build-bench is an end-to-end benchmark that evaluates the capability of large language models to repair build failures in cross-instruction set architecture (ISA) settings using a collection of 268 real-world failed packages and auxiliary tools for autonomous reasoning.

Papers using BUILD-BENCH (3)

Can Language Models Go Beyond Coding? Assessing the Capability of Language Models to Build Real-World Systems2025BuildBench: Benchmarking LLM Agents on Compiling Real-World Open-Source Software2025BuildBench: Benchmarking LLM Agents on Compiling Real-World Open-Source Software2025