FeijiangHan/PaperForge

An active paper-reading skill that reconstructs author reasoning, explains methods mechanistically, stress-tests assumptions, and generates follow-up research ideas.

What it solves

PaperForge provides a structured framework for reading and analyzing academic papers using LLMs. Instead of simple summaries, it helps researchers move beyond surface-level understanding to reconstruct the author's reasoning process, identify critical assumptions, and generate novel research ideas.

How it works

The project consists of a set of highly structured system prompts and "skills" designed for LLMs (like Claude or ChatGPT). When provided with a paper's link, title, or PDF, the agent follows a rigorous 12-step analysis pipeline:

  1. Problem Identification: Defines the research question and its value.
  2. Contextualization: Analyzes previous research gaps.
  3. Reasoning Reconstruction: Simulates the author's intuition and inspiration based on existing knowledge.
  4. Core Intuition: Distills the essence of the main idea.
  5. Methodology: Explains the pipeline with concrete examples.
  6. Theoretical Foundation: Breaks down mathematical derivations.
  7. Experimental Validation: Maps claims to specific experiments and results.
  8. Key Takeaways: Summarizes essential lessons.
  9. Assumption Testing: Identifies the most fragile assumptions.
  10. Reproduction: Suggests a minimal experiment for quick verification.
  11. Counter-argument: Designs potential counter-examples.
  12. Future Work: Proposes novel, non-incremental follow-up research ideas.

Who it’s for

Researchers and students in STEM fields (primarily) who want to deeply analyze academic literature and use LLMs to stress-test the logic of a paper rather than just summarizing it.

Highlights

  • Mechanistic Explanation: Focuses on how and why an idea was conceived, not just what the results were.
  • Critical Thinking: Includes specific steps for identifying weaknesses and designing counter-examples.
  • Human-like Style: Prompts the LLM to avoid generic AI phrasing and adopt a high-information-density style similar to researchers like Andrej Karpathy.
  • Customizable: Offers guidance on adapting the framework for social sciences and humanities.

Related

  • Project
    HKUSTDial/Supervisor-Skills

    A collection of structured AI skills and theoretical guides that distill a senior professor's research experience into executable prompts to help students navigate the research and paper-writing process.

  • Project
    meta-prompting/meta-prompting

    A prompting framework that uses structural templates instead of concrete examples to improve LLM reasoning, providing a formal mathematical foundation for automated prompt engineering.

  • Project
    dreadnode/rigging

    A lightweight LLM framework that simplifies production code integration by providing unified model access, structured Pydantic outputs, and type-safe prompt definitions.

  • Project
    shcherbak-ai/contextgem

    An open-source LLM framework for extracting structured data and insights from documents with precise source references and automatic justifications.

  • Project
    ianarawjo/ChainForge

    A visual data-flow environment for systematic prompt engineering, allowing users to battle-test prompts across multiple LLMs and model settings with built-in evaluation and visualization tools.