Write Paper

Scan the project directory for standard subdirectories:

• Result directories: results/, output/, data/, figures/, plots/
• Documentation: docs/, docs/derivations/, docs/discussions/
• Auto-detect detailed_derivations_*.md and discussion_*.md in docs directories or project root

If project_path is not provided, default to current working directory.

Stage 2: File Scanning and Analysis

Classify all files in result directories by type and analyze:

For each file, record: name, type, analysis summary, and path.

Stage 3: Background Document Loading

• Read derivations file (detailed_derivations_*.md) for theoretical foundations, mathematical proofs, derivation steps
• Read discussion file (discussion_*.md) for research motivation, key questions, research direction
• Read derivation plan (derivation_plan_*.md) for logical structure of theoretical framework
• Extract topic from document headers if not explicitly provided

Stage 4: Reference Search

If knowledge base is available:

• Search with topic query, retrieve up to 50 relevant references
• Pre-number results as [1], [2], ... for easy citation in section generation
• Format as context block providing reference titles and content summaries

Stage 5: Research Type Classification

Classify the research into: theoretical, computational, experimental, review, case_study, or mixed. Select appropriate section template based on classification.

See references/research-types.md for classification criteria and section templates for each research type.

Stage 6: Section-by-Section Generation

Generate each section individually with full context. For each section, provide:

• Role: distinguished professor and senior researcher
• All analyzed data files and their content
• Background documents (derivations, discussions)
• Knowledge base references (pre-numbered)
• Section-specific writing guidelines
• LaTeX formatting rules

See references/writing-guidelines.md for detailed per-section writing standards, information integration guidance, and quality criteria.

Stage 7: Citation Review

After generating all sections, review for citation completeness:

1. Claim extraction: Identify statements needing citations — factual claims, method attributions, theory references, result comparisons, background statements
2. Source search: For each uncited claim, search knowledge base and online literature (INSPIRE/arXiv) for relevant sources
3. Auto-insertion: Insert \cite{} markers or numbered references at appropriate locations
4. Report: Generate citation coverage statistics (% of claims with citations, uncited claims by importance)

Stage 8: Figure Handling

• Create paper/figures/ directory
• Copy all analyzed image files to paper/figures/
• Generate LaTeX figure blocks with \includegraphics, captions from vision analysis, and \label for cross-referencing:

\begin{figure}[htbp]
  \centering
  \includegraphics[width=0.8\textwidth]{figures/filename.png}
  \caption{Descriptive caption from vision analysis.}
  \label{fig:unique_label}
\end{figure}

Stage 9: LaTeX Assembly and Output

Assemble the complete document:

1. Generate LaTeX preamble (see template below)
2. Combine all sections with proper \section{} / \begin{abstract} wrappers
3. Append figure blocks at appropriate locations
4. Generate \begin{thebibliography} from collected references
5. Post-process: clean residual Markdown syntax, validate formulas, escape special characters
6. Save as paper/paper_[topic]_[date].tex
7. Generate paper/paper_[topic]_[date].bib with BibTeX entries

See references/latex-formatting.md for LaTeX syntax rules, formula standards, and post-processing rules.

LaTeX Preamble Template

\documentclass[11pt,a4paper]{article}
\usepackage[utf8]{inputenc}
\usepackage[T1]{fontenc}
\usepackage{amsmath}
\usepackage{amssymb}
\usepackage{amsthm}
\usepackage{graphicx}
\usepackage[margin=1in]{geometry}
\usepackage{caption}
\usepackage{booktabs}
\usepackage[colorlinks=true,linkcolor=blue,citecolor=blue,urlcolor=blue]{hyperref}

\title{Paper Title}
\author{Author Name}
\date{\today}

\begin{document}
\maketitle

Key Constraints

• Academic integrity: Never fabricate citations or references. Only cite sources that exist in the knowledge base or can be verified.
• Data fidelity: Use exact values from provided data files. Never invent numbers, statistics, or results. Quote actual row counts, means, standard deviations from real data.
• Writing structure: Write in coherent paragraphs (4-8 sentences each). No bullet-point papers — use academic prose. Smooth transitions between paragraphs.
• Pure LaTeX: No Markdown syntax in paper content. Use \textbf{} not **, \textit{} not *, proper \section{} not #.
• Section titles: No math expressions in \section{} or \subsection{} titles.
• Numbered references: Use [1], [2] format in text, compiled to BibTeX in output.

Reference Files

• references/research-types.md — Read when classifying research type or selecting paper structure template (Stage 5)
• references/writing-guidelines.md — Read when generating section content for detailed per-section standards (Stage 6)
• references/latex-formatting.md — Read when assembling LaTeX output for syntax rules, formula standards, and post-processing (Stage 9)