AutoCoder: The First Large Language Model to Surpass GPT-4 Turbo (April 2024) and GPT-4o in pass@1 on the Human Eval Benchmark Test (90.9% vs. 90.2%)

A novel method called AIEV-INSTRUCT has been introduced for creating high-quality code datasets, improving code generation capabilities. AutoCoder, trained with AIEV-INSTRUCT, surpassed GPT-4 Turbo in pass@1 on the Human Eval Benchmark Test (90.9% vs. 90.2%). The method reduces dependency on costly closed-source models and enhances the efficiency and accuracy of code generation tasks.