HKUSTDial/Supervisor-Skills

What it solves

It addresses the "last-mile problem" in academic research, where students often struggle to bridge the gap between general research guides and the practical execution of high-quality papers. It solves the lack of immediate, high-level guidance on idea evaluation, logical structuring, professional writing, and the nuances of top-tier conference submissions.

How it works

The project uses a dual-track architecture combining theoretical guides and executable AI skills:

• Handbook: A systematic theoretical framework covering paper quality evaluation, idea generation, writing methodologies, scientific plotting, and case studies of top-conference papers.
• AI Skills: A collection of structured prompts and skill files that "distill" a decade of senior research experience into executable instructions for LLMs (like Claude, GPT-4, DeepSeek). These skills cover the entire paper lifecycle, from initial idea scoring to final rebuttal guidance.

Who it’s for

Graduate students and early-career researchers who need a "virtual co-supervisor" to help them refine their research ideas, draft and polish papers, and navigate the peer-review process.

Highlights

• Comprehensive Lifecycle Support: Includes skills for idea evaluation, deep literature research, structured introduction drafting, and evidence-gated paper writing.
• Professional Polishing: Provides tools for language polishing that remove "AI-style" phrasing while remaining faithful to the original meaning.
• Reviewer Perspective: Features a pre-submission reviewer skill and a rebuttal guidance tool based on a conditional persuasion model.
• Visual Aid Tools: Includes a figure design advisor and a tool to reconstruct images into editable Draw.io files.
• Industry Integration: Deployed as core research skills within the Doubao Office Mode.