Combining both Mix and Hybrid mode gives best answer instead of each in individual mode

[Brahmananda-Rana]

When using LightRAG’s aquery_llm method with both Mix and Hybrid modes, and then synthesizing the results, the combined approach seems to provide better answers compared to using either mode individually.
Do you have any idea why LightRAG performs better when these modes are combined, given that both already handle graph and entity relationships?
We haven’t found any such recommendation in the official repository or online resources.

Below is the code:-

        hybrid_task = aquery_llm(query, mode = "hybrid")
        mix_task = aquery_llm(query, mode = "mix")
        hybrid_resp, mix_resp = await asyncio.gather(hybrid_task, mix_task)

        synthesis = f"""
        Question: {query}

        Hybrid context: {hybrid_resp}
        Mixed approach: {mix_resp}

        Instructions:
        - Provide a detailed and comprehensive answer to the question above.
        """
        answer= await llm_model_func(synthesis, deployment="gpt-5.1")