Adkid-Zephyr/oral-paper-skill

What it solves

Writing a high-quality academic paper that gets accepted as an "Oral" presentation at top AI conferences (like ICLR, ICML, and NeurIPS) is difficult. Researchers often struggle to articulate their contributions, organize evidence, and distinguish their work from previous research without relying on generic templates.

How it works

The project "distills" knowledge from 883 top-conference Oral papers. It analyzes their abstracts to extract common patterns, successful narrative strategies, and effective experimental designs. This distilled knowledge is then packaged as a "Skill" (a set of prompts and reference materials) that users can integrate into AI tools (like Claude or Codex) to receive tailored feedback on their own research ideas or drafts.

Who it’s for

Academic researchers and students in the AI field who want to improve their paper writing, refine their research narratives, and learn how to structure evidence to meet the standards of top-tier conferences.

Highlights

• Data-Driven Insights: Based on the analysis of 883 Oral papers from ICLR, ICML, and NeurIPS.
• Actionable Feedback: Provides specific suggestions for improvement based on the formula: original paper practice $\rightarrow$ why it is relevant $\rightarrow$ current draft status $\rightarrow$ specific modification.
• Focus on Research Tension: Helps authors clarify the specific limitations or phenomena that make a problem worth studying.
• Contribution Clarity: Guides users to define their actual contribution beyond just using words like "novel."
• Customizable Integration: Can be used as a standalone prompt or installed as a skill in supported AI assistants.