I am writing to seek assistance with an issue I am experiencing with the GPT-4 model in my Discord bot application. My bot is designed to respond to user queries, and I have integrated both the GPT-3.5-turbo and GPT-4 models for generating responses.
However, I am encountering a consistent issue with the GPT-4 model. Unlike the GPT-3.5-turbo model, which functions as expected, the GPT-4 model consistently fails to provide relevant answers to user queries. Regardless of the question asked, the GPT-4 model’s response is often unrelated or indicates a lack of information, even when the query is straightforward and should be within the model’s capability to answer.
For example, when asked about popular plugins for Minecraft, the GPT-4 model responds with: “I’m sorry, but the text doesn’t provide information on what the best plugins for Minecraft are.” This issue persists across various types of queries.
To troubleshoot this problem, I have tried the following steps without success:
- Ensuring that the bot correctly processes and sends user queries to the model.
- Verifying that the model receives the correct input format.
- Testing the bot with the GPT-3.5-turbo model, which works as expected.
Given these efforts, it seems the issue might be specific to the GPT-4 model’s integration or behavior. I would greatly appreciate any guidance or suggestions you could provide to resolve this issue. If additional information or logs are required, I am happy to provide them.