research-refine

To refine vague research ideas into concrete method plans, this Skill locks down the essential problem and uses an iterative two-model loop to generate an implementation-oriented methodology.

Iterative refinement for research methodology planning works by using a two-model loop with Claude and GPT-5.4 to continuously review and improve your method plan based on expert feedback.

To integrate modern primitives like LLMs and Diffusion into research plans, the methodology refinement process evaluates and incorporates these techniques naturally when they offer a clear advantage.

Yes, you can use this Skill to define minimal validation experiments for LLM reasoning. It encourages the smallest adequate mechanism and helps sketch out minimal validation experiments for top-venue readiness.

Yes, you need GPT-5.4 to iterate on your research method plan because the core refinement loop depends on a two-model architecture using Claude and GPT-5.4.

The best way to anchor a research problem to its core bottleneck is by locking down the essential problem and constraints during the initial refinement stages.