Igeekphone August 12 news, Microsoft Research recently jointly with the University of California, Berkeley, developed a new large language model (LLM) known as Gorilla, said in accuracy, flexibility and other API calls are better than GPT-4.
Microsoft Research says that even if the API documentation changes, the correct semantics and syntax are still generated when the API is called.
GPT-4 is certainly the most powerful LLM developed by OpenAI, which can handle various scenarios such as generating text, translating language, answering content, etc., but the API call is its short board, unable to understand the semantics of API calls, and can only generate text similar to API documents.
Gorilla LLM, developed by Microsoft Research, uses a technique called “retriever-aware training” to allow LLM to learn the semantics of API calls and generate semantically and syntactically correct text content.
In a recent study, Gorilla outperformed GPT-4 on a variety of API calling tasks.
GPT-4 has 85% accuracy in API calls, while Gorilla can achieve 95% accuracy.
Gorilla also has a skill that GPT-4 does not have to generate the right API calls for apis that do not have training datasets.