Publishing Papers

This guide covers the full workflow from experiment results to a published paper on arXiv. Hubify Labs handles drafting, review, compilation, and submission packaging.

Overview

Experiments → Claims Table → Outline → Draft → Review → Revise → Compile → Submit

The paper pipeline is agent-assisted at every step, but you maintain full editorial control.

Step 1: Create a Paper

1. Go to **Papers** in the sidebar
2. Click **New Paper**
3. Enter title, select template (`prd` for Physical Review D)
4. The Paper Lead generates an initial outline based on your lab's completed experiments


```bash
hubify paper create \
  --title "Observational Constraints on Bounce Cosmology" \
  --template prd
```

Step 2: Build the Claims Table

The claims table is the backbone of your paper. Every scientific assertion must be backed by evidence.

# View the claims table
hubify paper claims paper-1

#	Claim	Evidence	Confidence	Status
1	"Matter bounce predicts f_NL = -4.375"	EXP-031, Branch V	High	Verified
2	"NANOGrav signal consistent with bounce"	EXP-042, gamma=3.0	Medium	Verified
3	"Quintom-B favored at 2.3 sigma"	EXP-048, w0-wa MCMC	Medium	Under review

# Add a claim
hubify paper claim add paper-1 \
  --claim "SPHEREx can detect bounce f_NL at 4.7-12 sigma" \
  --evidence EXP-055 \
  --confidence high

Step 3: Draft Sections

Agents draft sections in parallel. You can direct which experiments feed into which sections:

# Have agents draft the full paper
hubify paper draft paper-1

# Or draft a specific section
hubify paper draft paper-1 --section results

# Review a section draft
hubify paper show paper-1 --section introduction

Step 4: Cross-Model Peer Review

Run a structured peer review using multiple AI models:

hubify paper review paper-1

The review checks:

• Accuracy — Are claims consistent with the data?
• Completeness — Are there missing analyses or citations?
• Clarity — Is the writing clear and well-structured?
• Overclaiming — Are conclusions justified by the evidence?
• Novelty — Does the paper clearly state what is new?

Review results appear as structured feedback with specific suggestions per section.

Step 5: Revise

Apply review feedback:

# Auto-apply review suggestions
hubify paper revise paper-1

# Apply specific suggestions
hubify paper revise paper-1 --apply "suggestion-3,suggestion-7"

# View revision history
hubify paper history paper-1

Each revision round is logged with a diff showing what changed and the rationale.

Step 6: Compile to PDF

Warning: LaTeX compilation requires texlive-publishers (for revtex4-2). This runs on a GPU pod, not locally.

# Compile
hubify paper compile paper-1

# View the PDF
hubify paper open paper-1

# Check for issues
hubify paper compile paper-1 --check

The compiler verifies:

• Zero undefined references
• All figures embedded (PDF size should be 15-25 MB for a paper with 10+ figures)
• Correct page count
• Bibliography completeness

Step 7: Package and Submit

# Create arXiv submission package
hubify paper package paper-1 --output submission.tar.gz

The package includes:

• Compiled PDF
• LaTeX source (.tex)
• Figures (all PNGs)
• Bibliography (.bib)
• Supplementary data (if any)

Upload the package to arxiv.org following their submission guidelines.

Tracking Readiness

Monitor paper progress in the readiness dashboard:

hubify paper status paper-1 --verbose

Paper: Observational Constraints on Bounce Cosmology
Status: In Review (Round 3)

Content:    ████████░░  80%  (6/7 sections drafted)
Figures:    ██████████  100% (11/11 placed)
Bibliography: █████████░  90%  (57/63 resolved)
Claims:     ██████████  100% (12/12 verified)
Compilation: ██████████  100% (0 errors)
Review:     ██████░░░░  60%  (3/5 rounds)

Overall Readiness: 88%

Best Practices

• Lock claims before drafting — do not write sections around unverified claims
• Run reviews after every major revision, not just at the end
• Compile frequently to catch LaTeX issues early
• Use revtex4-2, not aastex631
• Keep figures in the same directory as the .tex file