arXiv preprint

Free

Autonomous LLM-driven research from data to verifiable papers

FreeFree tier

Type

Open Source

Visit Website GitHub (40 stars)

About arXiv preprint

Data-to-paper is an automation platform that guides interacting LLM agents through a complete stepwise research process, from data to human-verifiable research papers. It programmatically back-traces information flow and allows human oversight. In autopilot mode, given annotated data alone, it can generate hypotheses, design research plans, write and debug analysis codes, generate and interpret results, and create traceable manuscripts. The system demonstrates autonomous generation of de novo quantitative insights, though novelty is limited. For simple research goals, fully-autonomous cycles can recapitulate peer-reviewed publications without major errors in about 80-90% of cases.

Key Features

Stepwise research process guided by interacting LLM agents

Programmatic back-tracing of information flow

Human oversight and interaction capabilities

Autopilot mode from annotated data to complete paper

Raises hypotheses and designs research plans

Writes, debugs, and executes analysis code

Generates and interprets results

Creates information-traceable manuscripts

Pros & Cons

Pros

Enables fully autonomous research for simple goals with high success rate (~80-90%)
Maintains traceability, transparency, and verifiability
Allows human co-piloting for complex tasks
Generates de novo quantitative insights from data

Cons

Research novelty is relatively limited
Human co-piloting becomes critical for complex research goals
May produce errors in complex scenarios without human oversight

Best For

Automated scientific discovery from structured dataRapid generation of draft research papers for simple analysesEnhancing reproducibility and transparency in computational research

FAQ

What is data-to-paper?

Data-to-paper is an automation platform that uses interacting LLM agents to perform a complete scientific research process, from data to a human-verifiable research paper.

Can data-to-paper run autonomously?

Yes, in autopilot mode, given annotated data, it can autonomously raise hypotheses, design experiments, write code, generate results, and produce a paper without human intervention.

What are the limitations of data-to-paper?

The research novelty is relatively limited, and for complex goals, human co-piloting is critical to ensure accuracy.