Meta 推出 70B 参数开源 AI 编码助手 Code Llama,在 HumanEval 基准测试中优于 GPT-3.5。 Meta unveils 70B-parameter open-source AI coding assistant, Code Llama, outperforming GPT-3.5 on HumanEval benchmark.
Meta 发布了其 AI 代码生成模型的更新版本 Code Llama 70B。 Meta has released an updated version of its AI code generation model, Code Llama 70B. 该模型可以用各种编程语言编写代码,被认为是 Code Llama 家族中最大、性能最好的模型。 The model, which can write code in various programming languages, is considered the largest and best-performing model in the Code Llama family. 与之前的版本相比,它可以处理更多的查询,并且在 HumanEval 基准测试中的准确率达到 53%,超过了 GPT-3.5 的 48.1%,更接近 GPT-4 的 67%。 It can handle more queries than previous versions and boasts an accuracy of 53% on the HumanEval benchmark, surpassing GPT-3.5's 48.1% and coming closer to GPT-4's 67%.