Papers

Hubify Labs treats paper writing as a first-class workflow, not an afterthought. The paper pipeline takes raw experiment results and produces arXiv-ready PDFs through a structured, agent-assisted process.

Paper Lifecycle

Results Collection

As experiments complete and pass QC, their results, figures, and metrics are tagged for inclusion in a paper. Agents organize results by paper section.

Outline & Claims

The Paper Lead drafts an outline and a claims table — a structured list of every scientific claim the paper makes, each linked to its supporting evidence.

Section Drafting

Agents draft individual sections (abstract, introduction, methods, results, discussion). Each section is reviewed by a different model before merging.

Cross-Model Peer Review

The full draft undergoes peer review from multiple AI models (Claude, GPT-4, Gemini). Reviews check scientific accuracy, logical flow, missing citations, and overstatements.

LaTeX Compilation

The manuscript is compiled using revtex4-2 (Physical Review D format). Figures are embedded, cross-references resolved, and the bibliography generated.

Revision Rounds

Based on peer review feedback, agents iterate on the draft. Each revision round is logged with a diff showing exactly what changed and why.

Submission

The final PDF, source files, and supplementary data are packaged for arXiv submission.

Paper Structure

Every paper in Hubify Labs follows a consistent structure:

\documentclass[aps,prd,twocolumn,superscriptaddress]{revtex4-2}

\begin{document}
\title{Your Paper Title}
\author{Your Name}
\affiliation{Your Institution}

\begin{abstract}
...
\end{abstract}

\maketitle

\section{Introduction}
\section{Methods}
\section{Results}
\section{Discussion}
\section{Conclusions}

\bibliography{references}
\end{document}

Warning: All papers use revtex4-2, not aastex631. Use \cite{}, not \citep/\citet. Use ruledtabular, not deluxetable.

Claims Table

The claims table is a living document that maps every assertion to evidence:

Claim	Evidence	Confidence	Status
"Matter bounce predicts f_NL = -4.375"	Branch V derivation, EXP-031	High	Verified
"Quintom-B favored at 2.3 sigma"	MCMC EXP-048, w0-wa chains	Medium	Under review

Agents cannot write a claim without linking it to an experiment or derivation. This prevents hallucinated results from entering the manuscript.

Readiness Tracking

Each paper has a readiness percentage that tracks progress across dimensions:

• Content — Are all sections drafted?
• Figures — Are all figures generated and placed?
• Bibliography — Are all citations resolved?
• Claims — Are all claims verified against evidence?
• Compilation — Does the LaTeX compile with zero errors?
• Review — Has the paper passed cross-model peer review?

Version History

Every change to a paper is versioned. The version history shows:

• What changed (diff)
• Why it changed (review feedback, new results)
• Who changed it (which agent)
• When it changed (timestamp)

CLI

# Create a new paper
hubify paper create --title "Spin-Torsion Cosmology" --template prd

# Check paper status
hubify paper status paper-1

# Compile to PDF
hubify paper compile paper-1

# Run peer review
hubify paper review paper-1

# List revision history
hubify paper history paper-1

Figure Management

Figures are automatically linked to experiments:

# List figures for a paper
hubify paper figures paper-1

# Add a figure from an experiment
hubify paper add-figure paper-1 --experiment EXP-054 --file posterior.png --caption "Posterior distribution"

All figures must be in the same directory as the .tex file (or symlinked) for LaTeX compilation to embed them correctly.