How do you deal with the hated trait of the gemini model, when he is asked to improve/supplement or change the code, then he writes the code, but skips the last details? He makes comments, but does not write the lines of code themselves until you tell him to write the code completely himself
Hi @USEnglish
In the initial prompt, clearly state that you want the complete code and ask the model not to stop or truncate the response midway.
If needed, you can also:
- Divide large prompts into smaller parts using chain prompting.
- Provide few-shot examples to guide the model.
Thanks.