DEV Community

Cover image for Runner H + Chat GPT: Automating Literature Review
mack-leod-b
mack-leod-b

Posted on

Runner H + Chat GPT: Automating Literature Review

This is a submission for the Runner H "AI Agent Prompting" Challenge


📚 What I Built

I used Runner H to analyze a research paper directly aligned with my project. I tasked the agent with generating a structured document that extracts and explains the key findings from the paper in a replicable format.


🎥 Demo

Document Structure Screenshot 1

Document Structure Screenshot 2


⚙️ How I Used Runner H

I started by describing what I wanted Runner H to do using a natural-language prompt:

*“I want you to read this paper comprehensively and identify key steps that they used to achieve their results. After identifying those, give me a detailed explanation of their approach and explain it to me in such a way that I can replicate it. Also give me a summary of everything as well. Create a @tool:Google Docs document. Write into the Google Docs document. The structure should be this way:

  1. Summary of the Paper
  2. Phases of the Project
  3. Key Steps Taken to Come Up with the Solution
  4. Explanation of the Key Steps They Used (with Code)
  5. Suggestions on What Next to Do Since I’m Working on a Similar Project”*

Then I asked ChatGPT to refine this prompt for clarity and effectiveness with Runner H. Here's the structured response I received:


🧠 Final Runner H Prompt Structure

1. Summary of the Paper

  • A concise summary (150–300 words) capturing the objective, methods, datasets used, key findings, and conclusions.
  • Highlight the problem they aimed to solve and the core contribution of the paper.

2. Phases of the Project

  • Break down the project into logical phases or stages (e.g., data collection, preprocessing, modeling, evaluation).
  • Label and describe each phase clearly.

3. Key Steps Taken to Achieve the Results

  • Identify and list the critical decisions, methods, algorithms, or tools used.
  • Include any preprocessing techniques, feature engineering, model selection, tuning, evaluation metrics, and tools (e.g., Python libraries, hardware used).

4. Detailed Explanation & Replication Guide (with Code)

For each key step:

  • Explain what was done and why.
  • Provide code snippets (in Python or pseudocode if needed) that replicate the methodology.
  • Note external dependencies or datasets and where to get them.

5. Suggestions for Next Steps in My Own Project

Given that I’m working on a similar project (e.g., predicting bilirubin levels in infants using image data):

  • Suggest how to adapt or extend the paper’s methods.
  • Mention potential limitations or challenges in the original paper.
  • Offer ideas for improvement, modern alternatives, or follow-up experiments.

📝 Additional Notes

  • Assume intermediate-level Python and ML knowledge.
  • Use diagrams or tables where helpful.
  • If parts of the paper are unclear, make educated guesses based on similar research.

💡 Use Case & Impact

This use of Runner H significantly reduces the time and effort typically spent on literature reviews in research. It not only automates the summarization but also creates a reproducible, code-friendly interpretation that can serve as a launchpad for practical experimentation.


❤️ Social Love

If you're a researcher or student working on a machine learning or scientific project, you’ll find this approach incredibly useful!

Check Out this on DEV Community.
Runner H Challenge

Here is what I used the tool for: Automated my Literature Review https://t.co/qjU7qFuTWj

— Ben_z (@crypticBen_z) June 24, 2025

Image description

Top comments (0)