The community member is experiencing issues with the LlamaIndex library, where the LLM (Hugging Face H4/zephyr-7b-alpha) works fine with regular requests but sometimes starts making and asking its own questions after getting results when using the QueryEngine. Another community member suggests that this could be an issue with the smaller LLM model, which is more prone to hallucinations and random outputs.
hey, i have some trouble with llamaindex, llm works fine with regular request, but with queryengine it sometimes starts making and asking own questions after getting result, what could it be?
