Skip to main content
Back

AgentAlpha's Idea2Paper Turns Rough Research Ideas Into Submission-Ready Papers

AgentAlphaAGI open-sources Idea2Paper, a modular agent framework that turns a raw research idea into a structured scientific manuscript using a pre-built knowledge graph

Pro

  • What it is: Idea2Paper is an open-source (MIT) Python framework that converts a raw research idea into a structured scientific narrative skeleton.
  • Core innovation: Moves literature understanding offline into a pre-built knowledge graph (13K ICLR/NeurIPS papers), replacing expensive runtime reasoning with graph retrieval.
  • How it works: Three-path retrieval (idea, domain, paper level) + anchored multi-agent review that scores outputs against real peer-review anchors for reproducible, auditable results.
  • Current scope: Produces a story skeleton (title, abstract, problem, method, contributions, experiments) — full manuscript generation is not yet automated.
  • New module: Story2Proposal extends the pipeline toward full structured manuscript generation using a contract-governed multi-agent loop.
  • Getting started: Prebuilt dataset on Hugging Face; run with one command after setting your LLM API key.

Idea2Paper is an open-source agent framework from the AgentAlpha team that takes a raw research idea and produces a structured, submission-ready scientific narrative. The repo has picked up over 1,300 stars and 100+ forks, and it ships with a local web UI, a prebuilt dataset on Hugging Face, and a companion arXiv paper detailing the core module.

The problem every AI research agent has

Autonomous scientific discovery with LLM-based agents has made real progress, but existing systems largely rely on runtime-centric execution, repeatedly reading, summarizing, and reasoning over large volumes of scientific literature on the fly. That on-the-spot strategy incurs high computational cost, suffers from context window limitations, and often leads to brittle reasoning and hallucination. In practice, this means a single run of an end-to-end research pipeline can take anywhere from several hours to over 15 hours just to go from ideation to experimentation.

Idea2Paper is an end-to-end research agent framework that explicitly decomposes scientific research into structured phases and identifies critical bottlenecks that hinder the transformation of raw ideas into coherent, submission-ready academic narratives. One of the most fundamental yet underexplored challenges it targets is research paradigm generation: the process of converting an underspecified research idea into a logically consistent, academically grounded research story.

Offline first, runtime second

Rather than building a monolithic end-to-end writing system, Idea2Paper adopts a modular design that prioritizes targeted engineering submodules. The first and core submodule is Idea2Story, introduced to directly address research paradigm generation. Idea2Story focuses on transforming underspecified research ideas into complete, coherent, and submission-ready scientific narrative skeletons.

The key architectural bet is moving literature understanding offline. Idea2Story is a pre-computation-driven framework that shifts literature understanding from online reasoning to offline knowledge construction. It continuously collects peer-reviewed papers together with their review feedback, extracts core methodological units, composes reusable research patterns, and organizes them into a structured methodological knowledge graph.