watericetangcw / academic-research-graph

Academic Research Graph Skill

简体中文

Turn one paper into a living research map.
Trace where an idea came from, how it evolved, who challenged it, and what the code actually does.

Example instruction: "Research the concept Transformer and paper Attention Is All You Need with @academic-research-graph."

Give the Skill a paper, academic concept, benchmark, claim, or paper-associated GitHub repository. It researches the surrounding literature and returns an interactive graph you can actually explore—not just a long list of citations.

See the field, not just the bibliography

Follow the main intellectual lineage

See which earlier work supplied the key idea, which papers genuinely extended it, and which branches became important later.

Understand papers you have never read

Each important node explains the paper for someone who knows the field but has not opened the PDF: background, problem, method, evidence, significance, and limitations.

Find real disagreements

Conflict edges highlight criticism, contradictory findings, failed reproductions, incompatible assumptions, and cases where two papers only appear to disagree because they use different data or evaluation protocols.

Verify benchmark adoption

The graph distinguishes a paper that merely cites a benchmark from one that actually evaluates on it, loads its data, or uses its official code.

Connect the paper to the code

Start from a GitHub repository, resolve the associated paper, inspect the implementation at a pinned commit, and compare the code with the method described in the paper.

Keep growing the graph

Select any existing node, launch a new research branch, and merge the reviewed findings back into the original graph without rebuilding everything.

What you get

The generated page can show:

• clickable paper nodes and relationship edges;
• preprint, conference, journal, and publication status;
• sourced top-venue badges;
• readable paper guides and metadata links;
• main-lineage and controversy highlighting;
• optional representative figures for key papers;
• Chinese, English, or bilingual content;
• “latest expansion” highlighting only after a real incremental update.

Live example: Open the interactive research graph

Start with almost anything

• a paper URL, DOI, arXiv ID, title, or PDF;
• a concept such as Vision Transformer;
• a benchmark or dataset;
• a concrete academic claim or controversy;
• a paper-associated GitHub repository;
• any node in a graph you already built.

Use it in three steps

1. Install it in ChatGPT or DeepSeek Harness using the instructions below.
2. Describe what you want to investigate, including the focus and output language.
3. Explore the generated graph, then continue from any node when a branch deserves deeper research.

Example requests

Research arXiv:2406.05720. Trace its important origins and meaningful
follow-up work, identify major disagreements, and verify which papers
actually used its benchmark. Produce the graph in English.

Build a Chinese research graph for Vision Transformer. Explain every core
paper for a field-familiar reader who has not read it, and highlight the
main lineage from Transformer to ViT and its most important descendants.

Start from this GitHub repository, identify the associated paper, compare
the implementation with the paper, and map its origins, follow-up work,
and criticism.

Continue from the DeiT node in the current graph. Investigate the data-
efficiency branch and merge the reviewed results back into the graph.

Language support

Choose one output mode:

output_language: zh-CN      # Simplified Chinese
output_language: en         # English
output_language: bilingual  # Chinese + English

Paper titles, author names, venue names, identifiers, benchmark names, metrics, equations, and code symbols remain in their canonical form so they stay searchable.

Download and install

ChatGPT

1. Download the latest archive from https://github.com/watericetangcw/academic-research-graph/releases/latest.
2. Open ChatGPT Skills.
3. Import the ZIP or extracted Skill folder.
4. Use SKILL.md as the Skill entry point when your host asks for one.

DeepSeek Harness

This repository also ships as a native DeepSeek Harness profile bundle. The bundle format targets @deepseek-ai/dsh 0.1.0-rc.5.

Prerequisites:

• Node.js ^22.19.0 or >=24.0.0;
• pnpm available on PATH (corepack enable can provide it).

Install the plugin into the default web profile, then start Harness:

corepack enable
npx @deepseek-ai/dsh plugin --profile web add github:watericetangcw/academic-research-graph
npx @deepseek-ai/dsh web

After Harness starts, enter /academic-research-graph followed by your research request. Harness may also load the Skill automatically when the request matches its description.

Plugins are profile-scoped. Replace web with headless or another profile name when that is the profile you run. Append #v0.2.4 to the GitHub package spec if you want to pin this release.

Inspect, update, or remove the installed bundle with:

npx @deepseek-ai/dsh --profile web --dump-config
npx @deepseek-ai/dsh plugin --profile web update dsh-academic-research-graph
npx @deepseek-ai/dsh plugin --profile web remove dsh-academic-research-graph

If dsh is already installed as a command, you can use dsh instead of npx @deepseek-ai/dsh.

For developers

The main README focuses on the experience delivered to research users. Architecture, schemas, scripts, local rendering, graph patches, adapter contracts, repository structure, asset rules, and contribution guidance live here:

• Developer Guide
• 开发者指南（中文）
• Skill workflow
• Detailed design

License

This project is licensed under the MIT License.

Citation

If you use Academic Research Graph Skill in research or academic work, please cite the project using the metadata in CITATION.cff.