Replies: 1 comment 2 replies
-
You're right. Code Interpreter internals aren't fully disclosed (e.g. which prompts they're using). They likely also have additional logic implemented around code execution and execution parsing (and feeding that into the context window for consecutive generation). Furthermore, they likely also use a different model as hinted at by the |
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
It seems when I input the same questions in this platform and the gpt4 code interpreter, the answers are far different. Can I understand that as this project adds some prompts based on gpt 4 or 3.5, but the code interpreter has some more built-in that only Openai knows?
Beta Was this translation helpful? Give feedback.
All reactions