Related Work Co-pilot

An AI-powered assistant for drafting related work sections, combining large language models and arXiv with a human-in-the-loop approach for structured, iterative refinement.

📂 View the code on GitHub 🎬 Watch Demo Video

🎯 Motivation

As a researcher, I often found myself overwhelmed while writing the related work section — juggling keywords, chasing arXiv threads, and trying to structure something meaningful.

Tools like OpenAI’s Deep Research are fascinating, but I craved more control, more transparency, and a process that didn’t write for me, but with me.

That’s why I tried building Related Work Co-pilot — an AI-powered assistant that lets you drive every decision: from finding papers to shaping your final draft.

⚙️ This project is a personal response to the automation trend: proving that AI doesn’t have to replace us — it can empower us, especially in the most nuanced parts of research writing.

👨‍🔬 What It Is

Related Work Co-pilot helps you:

  • ✅ Define your topic with the help of AI-suggested, editable keywords
  • 📚 Find relevant arXiv papers through RAG-style iterative refinement
  • 🧠 Categorize those papers into meaningful sections using LLMs
  • ✍️ Draft a related work section — in Text or LaTeX — with your guidance
  • 🔁 Refine that draft interactively until it truly reflects your voice
  • 🧾 Export your citations in BibTeX, ready for submission

Unlike tools that push out paragraphs with minimal oversight, this co-pilot keeps you in control — step-by-step with unlimted iterative refinements per step.

✨ Why This Matters

🚫 Most research-assist tools are “black boxes.”
✅ This one invites you to steer the wheel.

I followed my own workflow for conducting literature review. This tool halves the time we spend searching and drafting — and gives better structure and citation accuracy.

🧭 Workflow Overview

📝 1. Define Your Topic & Keywords

  • Add your topic
  • Enter or AI-suggest keywords
  • Provide anchor papers or DOIs
  • Click Find Papers

📚 2. Explore & Select Papers

  • Search results from arXiv
  • Browse abstracts, select papers
  • Use AI to refine the search based on selections
  • Iterate until satisfied

🗂️ 3. Structure with AI Help

  • Choose number of sections
  • LLM proposes titles & descriptions
  • Automatically assigns papers
  • You revise everything if needed

✍️ 4. Draft & Refine

  • Draft in LaTeX or plain text
  • Ask for refinements:
    • “Make section 1 more critical”
    • “Expand on paper X”
  • Restore earlier versions

📄 5. Export & Cite

  • Copy or download the draft
  • Get BibTeX entries for selected papers
  • Ready for integration into your paper

🖼️ Interface Preview

Designed for transparency and flexibility — choose papers, organize your structure, and edit drafts iteratively.

🧪 Technologies Used

  • 🧠 AI Providers: Google Gemini & OpenAI
  • 📚 Paper Data: arXiv API
  • 🛠️ Frontend: React, Tailwind CSS
  • 🚀 Build: Vite, Node.js

Note: The components of this project were built with the assistance of Google AI Studio.