API is superior to GPT-4 in calling, and Microsoft developed Gorilla LLM.

CTOnews.com, August 12, Microsoft Research, in conjunction with the University of California, Berkeley, has recently developed a new large language model (LLM) called Gorilla, which is superior to GPT-4 in terms of accuracy and flexibility in calling API.

Microsoft Research says that even if the API document changes, it can still generate the correct semantics and syntax when calling API.

It is true that GPT-4 is the most powerful LLM developed by OpenAI, which can control various scenarios such as generating text, translation language, response content, etc., but the aspect of API call is its deficiency, so it cannot understand the semantics of API call, and can only generate text similar to API documents.

Gorilla LLM, developed by Microsoft Research Institute, uses a technology called Retrieval Awareness training (retriever-aware training) to let LLM learn the semantics of API calls and generate semantically and grammatically correct text content.

In a recent study, Gorilla outperformed GPT-4 on various API invocation tasks.

The accuracy of GPT-4 calls in API is 85%, while that of Gorilla is 95%.

In addition, Gorilla has skills that GPT-4 does not have to generate the correct API calls for API without a training dataset.

CTOnews.com is here to attach a link to a detailed blog post, which interested users can read in depth.