Gemini 的 DeepResearsh 输出的报告和 ChatGPT 的相比信息密度低很多，读起来感觉很累。 
两家我输入的内容是一样的，为了排除中文的影响，都是英文的输入。 
I'm currently fine-tuning a large language model (LLM) on a proprietary codebase. The fine-tuning process completed without technical issues, but the resulting model performs very poorly—its responses are often irrelevant, even when prompted with questions that were directly taken from the training data.

Th...