AutoCodeBench

AutoCodeBench is a large-scale code generation benchmark containing 3,920 problems evenly distributed across 20 programming languages, used to evaluate the code generation capabilities of large language models without relying on manual annotations.

Papers using AutoCodeBench (2)

AutoCodeBench: Large Language Models are Automatic Code Benchmark Generators2025AutoCodeBench: Large Language Models are Automatic Code Benchmark Generators2025