微软探讨大型语言模型在医疗领域错误检测与修正的应用，并推出了名为MEDEC的全新基准测试集。然而，这篇论文同时也估计了市面上多个商业模型的规模，研究团队认为OpenAI GPT-4o-mini仅约80亿参数，远小于外界先前的预期。

标题：Anthropic研究发现：打错字可助攻解锁GPT-4等AI模型秘密 ...

Users can type the @ symbol to choose a GPT ... Paying users of ChatGPT can make a maximum of 25 GPT-4 queries every three hours, IEEE pointed out. In some jobs, the AI may remove the need ...

OpenAI has released GPT-4, the latest version of its hugely popular artificial intelligence chatbot ChatGPT. The new model can respond to images - providing recipe suggestions from photos of ...