humaneval

#自然语言处理#  We introduced a new model designed for the Code generation task. Its test accuracy on the HumanEval base dataset surpasses that of GPT-4 Turbo (April 2024) and GPT-4o.

  Run evaluation on LLMs using human-eval benchmark