Improving the quality of responses can be approached in the following three dimensions:
Model Selection Strategy
- Preference for models with larger parameters in specialized areas (e.g. Llama 3-70B is preferred over version 7B)
- For creative tasks, try Gemini-1.5 or Claude-3!
Function Combination Tips
- Factual Questions Turn on Web Search to Ensure Timeliness of Information
- Complex problem split into multiple rounds of dialog (model parallel processing context)
Cue word optimization
- Add a role setting (e.g. "You are a senior Python engineer")
- Clarify output formatting requirements ("compare advantages and disadvantages with tables")
Supplementary program: developers can modifylib/configs/modelConfig.tsAdjustment of model parameters such as temperature values controls output randomness.
This answer comes from the articleOpen-Fiesta: an open source tool for chatting with multiple AI macromodels at onceThe
May not be reproduced without permission:AI productivity tools " How should I optimize the experience of using the model if I encounter poor quality answers?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil