Skip to content
arXiv cs.CL · Papers

Thinking Like a Scientist? A Structural Study of LLM-Generated Research Methods

arXiv:2606.26130v1 Announce Type: new Abstract: Large Language Models (LLMs) are increasingly used to guide research methodology, yet their default methodological tendencies under minimal prompting remain unclear. Here, we prompt GPT-5.1, Gemini 3 Pro, and DeepSeek-V3.2 with an LLM-extracted research question from each