Section Writing Agent

Section Writing Agent (Step 4)

Faithful implementation of the Section Writing Agent from PaperOrchestra (Song et al., 2026, arXiv:2604.05018, §4 Step 4, App. F.1 pp. 47–49).

Cost: ONE LLM call (App. B: "Section Writing Agent (1 call): A single, comprehensive multimodal call to draft and compile the complete LaTeX manuscript"). Do NOT split this into per-section calls — the paper explicitly designs it as one comprehensive call so the model can maintain global coherence across sections.

Inputs

• workspace/outline.json — the master plan
• workspace/inputs/idea.md — technical details
• workspace/inputs/experimental_log.md — raw data for tables and qualitative analysis
• workspace/drafts/intro_relwork.tex — the template **with Intro + Related

Work already filled in by Step 3**. This is your starting point. The preamble, package list, style, and the two pre-filled sections must be preserved verbatim.

• workspace/citation_pool.json — the citation map ({key, title, abstract}

for each verified paper)

• workspace/refs.bib — the BibTeX file
• workspace/inputs/conference_guidelines.md — formatting rules
• workspace/figures/ — the actual PNG files from Step 2 (used as

multimodal vision input!)

• workspace/figures/captions.json — caption text per figure_id
• workspace/tex_profile.json — TeX package availability flags (written by

check_tex_packages.py at Step 0). Read this before generating any LaTeX. It tells you which packages are installed so you select the right cross-reference pattern, font packages, etc. before you write — not after you try to compile.

Output

• workspace/drafts/paper.tex — the complete LaTeX paper, with all sections

filled. The Step 5 Refinement Agent will iterate on this file.

How to do it

0.5. Read tex_profile.json and select LaTeX patterns

Before composing the prompt, read workspace/tex_profile.json and apply these rules to every LaTeX choice in the generated paper:

If tex_profile.json does not exist (old workspace), default to the safe fallback column (no cleveref, no nicefrac, no microtype, no T1 fontenc).

1. Pre-extract metrics from the experimental log

Run the deterministic helper:

python skills/section-writing-agent/scripts/extract_metrics.py \
    --log workspace/inputs/experimental_log.md \
    --out workspace/metrics.json

This parses the ## 2. Raw Numeric Data section's markdown tables into structured JSON. The Section Writing Agent uses this to construct LaTeX booktabs tables without re-deriving values from raw text. Read references/latex-table-patterns.md for the booktabs conventions.

2. Compose the prompt and make ONE multimodal call

Load references/prompt.md (verbatim Section Writing Agent prompt from App. F.1). Prepend the Anti-Leakage Prompt from ../paper-orchestra/references/anti-leakage-prompt.md.

The user message contains:

• outline.json — full content
• idea.md — full content
• experimental_log.md — full content (tables AND prose)
• intro_relwork.tex — full content (this becomes template.tex for the prompt)
• citation_pool.json — full content (becomes citation_map.json)
• conference_guidelines.md — full content
• figures_list — array of {figure_id, filename, caption} from

captions.json and the file listing

• The actual figure PNGs as multimodal image inputs, so the model can

visually inspect them and write accurate descriptions / refer to them correctly in the prose.

If your host LLM has no vision input, fall back to text-only mode: pass the captions in captions.json as descriptions and tell the agent it cannot see the images directly. Quality drops noticeably (the paper notes that visual grounding measurably improves figure-text alignment), but the pipeline still completes.

3. Save the output

The agent's response is wrapped in \\\latex ... \\\` fences. Extract the LaTeX code and save to workspace/drafts/paper.tex`.

4. Run the deterministic gates

# Orphan citation gate: every \cite{KEY} must exist in refs.bib
python skills/section-writing-agent/scripts/orphan_cite_gate.py \
    workspace/drafts/paper.tex workspace/refs.bib

# Latex sanity: matched braces, matched begin/end, no unescaped specials
python skills/section-writing-agent/scripts/latex_sanity.py \
    workspace/drafts/paper.tex

# Anti-leakage post-check: no author names, emails, affiliations
python skills/paper-orchestra/scripts/anti_leakage_check.py \
    workspace/drafts/paper.tex

If any gate fails, re-prompt the writing call with the gate's error report appended to the user message and ask the agent to fix the specific issues. Do NOT try to fix the gate violations by hand — the model needs to see its own mistakes.

Critical rules from the prompt

These are excerpted from references/prompt.md (App. F.1, pp. 47-49). The host agent MUST honor them on the writing call:

Existing-content preservation

• DO NOT modify the text, style, or content of sections that are already

filled in intro_relwork.tex. Preserve Intro + Related Work verbatim.

• Keep the preamble (packages, document class, style) exactly as is.
• Come up with a good title if one is missing. Fill author names if missing

(but the Anti-Leakage Prompt says not to invent real ones — use a placeholder like "Anonymous Authors" for double-blind).

Data and tables

• Build LaTeX tables for the experimental results.
• Extract numeric values directly from experimental_log.md. **Do not

hallucinate numbers** — use the exact values in the log.

• Use the booktabs package format: \toprule, \midrule, \bottomrule.
• All tables must appear before the Conclusion section, unless they are

explicitly placed in an Appendix.

Citations

• The outline.json provides citation_hints per subsection. For each hint,

find the matching key in citation_pool.json (by title or content) and use that exact key in \cite{...}.

• Use ONLY keys from `refs.bib`. Inventing or guessing keys violates the

Lit Review Agent's verified pool.

• Read the abstract from citation_pool.json for the papers you cite.

Use the abstract context to write specific, accurate sentences about those works — not generic "[A, B] proposed methods for X".

Writing content

• Write the missing sections following outline.json's section_plan

structure exactly. Hierarchy rule: if 4.1 exists, 4.2 must exist.

• Use formal mathematical equations, notations, and definitions where

appropriate AND directly supported by idea.md or experimental_log.md. Do not hallucinate math. Do not use complex math just for the sake of it.

• Always provide detailed ablation studies and qualitative analysis of the

experimental results: what worked, what does not, and why.

• Optional: discuss limitations and future work at the end.
• If you put anything in the Appendix, the Appendix section appears AFTER

the References section, on a fresh new page.

Figures and visual fidelity

• You are being given the actual image files of the figures. You MUST

describe them faithfully and accurately. Do NOT hallucinate interpretations that contradict the visual evidence in the plots.

• Use ALL of the figures provided in figures/. Use the exact filenames

including extensions (e.g., .png) in your \includegraphics commands.

• DO NOT merge or group multiple figures into one display.
• If the paper is in a 2-column format, prefer single-column figures

(\begin{figure}) unless they are very wide.

• All figures must appear before the Conclusion section, unless explicitly

in the Appendix.

• Refine the captions if necessary, but they are already provided in

captions.json and should generally be used as-is.

• Do NOT include "Figure X" in the caption text — LaTeX handles numbering.

Style

• Adopt the tone of a top-tier ML conference paper: dense, objective,

technical.

• Match the indentation and spacing style of the original template.tex.

Do not change the overall LaTeX style.

LaTeX integrity

• The output must compile flawlessly out-of-the-box.
• All \begin{X} must match a \end{X} (e.g., \begin{figure*} must be

closed with \end{figure*}, not \end{figure}).

• DO NOT change \usepackage[capitalize]{cleveref} to

\usepackage[capitalize]{cleverref} — there is no cleverref.sty.

• Always emit `\clearpage` immediately before `\bibliographystyle{...}`.

Without it, figures deferred by LaTeX's float algorithm will appear inside or after the References section — a hard-to-spot layout defect that only shows up in the compiled PDF. \clearpage forces all pending floats to be output before the bibliography starts. See references/latex-table-patterns.md for details.

• Cross-references: prefer Figure~\ref{fig:X} and Table~\ref{tab:Y}

over bare \ref{fig:X}. This is necessary when cleveref is unavailable and produces readable prose in all cases. Use \cref{...} only when cleveref.sty is confirmed present.

Output format

• Wrap the full updated template.tex in \\\latex ... \\\``.
• The previously empty sections should now be filled.
• Previously filled sections (Intro, Related Work) should remain mostly

untouched; only adjust for consistency purposes.

Resources

• references/prompt.md — verbatim Section Writing Agent prompt from App. F.1
• references/latex-table-patterns.md — booktabs rules + table-from-log examples
• references/figure-integration.md — \includegraphics, 2-column handling, placement
• scripts/extract_metrics.py — markdown tables in experimental_log → JSON
• scripts/latex_sanity.py — unmatched braces, env mismatches, specials
• scripts/orphan_cite_gate.py — every \cite{KEY} exists in refs.bib

Figure Integration

Conventions for including figures in the LaTeX paper, per the Section Writing Agent prompt (App. F.1 p.48, item 5 "Figures and Visual Fidelity").

Where the figures live

After Step 2 (Plotting Agent), figures are at:

workspace/figures/
├── fig_framework_overview.png
├── fig_main_results.png
├── fig_ablation_temperature.png
└── captions.json

The Section Writing Agent must reference them with the exact filenames, including the .png extension:

\begin{figure}[t]
\centering
\includegraphics[width=0.95\linewidth]{figures/fig_framework_overview.png}
\caption{Overview of the proposed pipeline. Raw video frames flow into the
frozen SAM encoder; aligned audio cues are projected through the temporal
modality fusion layer before being injected into the mask decoder.}
\label{fig:framework_overview}
\end{figure}

Single-column vs full-width

The prompt is explicit: in 2-column conference templates, prefer \begin{figure} (single-column) unless the figure is very wide. Use \begin{figure*} only for cross-column figures.

Caption placement

For figures, \caption goes AFTER \includegraphics:

\begin{figure}[t]
\centering
\includegraphics[width=0.95\linewidth]{figures/fig_main_results.png}
\caption{Comparison of three temporal-attention variants on the Ref-AVS
Seen split. Bars show Jaccard index; error bars are 95% confidence
intervals over five seeds.}
\label{fig:main_results}
\end{figure}

(For tables, \caption goes BEFORE \begin{tabular}. Different conventions; both correct in their respective contexts.)

Caption text

• Pull from workspace/figures/captions.json keyed by figure_id.
• Do not include Figure N: in the caption text. LaTeX adds the prefix

via \caption.

• Plain text only. No markdown bold/italic.
• 1-3 sentences. State what the figure shows AND why (the takeaway).

Reference in prose

Every figure must be referenced in the prose:

As shown in Figure~\ref{fig:main_results}, our method outperforms both
baselines across all five splits.

Use ~ (non-breaking space) before \ref{...}. Use Figure capitalized when starting a sentence; lowercase figure mid-sentence.

DO NOT merge figures

The prompt forbids combining multiple figures into one display. Each figure_id from the outline corresponds to exactly one \begin{figure} environment.

All figures before Conclusion

Per the prompt: "all figures must appear before the Conclusion section, unless they are placed in an Appendix." If you have a figure that contextually belongs in the Appendix, move it there explicitly — do not leave it floating after the Conclusion in the main body.

Multi-panel figures

If a figure has multiple panels (a, b, c) — e.g., a 1×3 grid showing ablations across three settings — the panels are part of the same PNG file (rendered by Step 2). Use a single \includegraphics and reference sub-panels in the caption text:

\caption{Ablation study. (a) Effect of temperature. (b) Effect of dropout.
(c) Effect of layer count. Bars show validation accuracy on the Seen split.}

If you really need separate sub-figure environments (subcaption package), that's allowed but adds complexity — prefer single-PNG multi-panel from Step 2.

LaTeX Table Patterns

Conventions for building LaTeX tables from experimental_log.md raw numeric data, per the Section Writing Agent prompt requirements (App. F.1 p.47, item 2 "Data & Tables").

Required: booktabs

Always use the booktabs package. The preamble in a typical conference template already includes it; if not, add:

\usepackage{booktabs}

Three rules only

Booktabs uses only three horizontal rules: \toprule, \midrule, \bottomrule. No \hline. No vertical bars.

\begin{table}[t]
\centering
\caption{Comparison of methods on Dataset X.}
\label{tab:main_results}
\begin{tabular}{lccc}
\toprule
Method      & Accuracy & F1 & Latency (ms) \\
\midrule
Baseline    & 78.2     & 0.79 & 12.3 \\
\textbf{Ours} & \textbf{85.4} & \textbf{0.87} & \textbf{8.1} \\
\bottomrule
\end{tabular}
\end{table}

From experimental_log markdown table → LaTeX

experimental_log.md contains tables in plain markdown:

## 2. Raw Numeric Data

### Table 1: Performance comparison on Dataset X

| Method   | Accuracy | F1   | Latency (ms) |
|----------|----------|------|--------------|
| Baseline | 78.2     | 0.79 | 12.3         |
| Ours-S   | 82.1     | 0.83 | 9.4          |
| Ours-L   | 85.4     | 0.87 | 8.1          |

The extract_metrics.py helper parses these into JSON:

{
  "tables": [
    {
      "label": "Performance comparison on Dataset X",
      "headers": ["Method", "Accuracy", "F1", "Latency (ms)"],
      "rows": [
        ["Baseline", "78.2", "0.79", "12.3"],
        ["Ours-S", "82.1", "0.83", "9.4"],
        ["Ours-L", "85.4", "0.87", "8.1"]
      ]
    }
  ]
}

The Section Writing Agent then converts each entry to a table environment verbatim. Important rules from the prompt:

• Do not hallucinate numbers. Copy the exact values from

extract_metrics.py's output.

• Bold the best result in each column (the convention for top-tier ML

papers).

• Use `\multicolumn{N}{c}{...}` for grouped headers when the table has

metric families (e.g., "Seen (J%)", "Seen F", "Unseen (J%)", "Unseen F").

• Right-align numeric columns with r, left-align text columns with l.

Use c only for narrow centered identifiers.

• Use `\textbf{...}` for bold, never **...** (markdown).

Wide tables (2-column conference templates)

For tables that don't fit single-column width, use table* and tabular* or tabularx:

\begin{table*}[t]
\centering
\caption{Ablation across all 6 components on 4 splits.}
\label{tab:ablation}
\begin{tabular}{lcccccc}
\toprule
Variant       & Seen J & Seen F & Unseen J & Unseen F & Mix J & Mix F \\
\midrule
Full          & 43.43  & 0.568  & 54.58    & 0.664    & 49.01 & 0.616 \\
- TB          & 33.05  & 0.507  & 50.48    & 0.657    & 41.77 & 0.582 \\
- TMFL        & 40.35  & 0.579  & 45.54    & 0.627    & 42.95 & 0.603 \\
\bottomrule
\end{tabular}
\end{table*}

The closing \end{table*} must match the opening \begin{table*}. The latex_sanity.py script catches mismatches.

Caption placement

\begin{table}[t]
\centering
\caption{Caption text here.}        % BEFORE the tabular for tables
\label{tab:my_label}
\begin{tabular}{...}
...
\end{tabular}
\end{table}

(For figures, \caption goes AFTER \includegraphics, not before. See figure-integration.md.)

Common pitfalls

Figures floating into or after the References section

This is the most common final-layout bug. When many figures appear in the Experiments section and the bibliography is near the end, LaTeX cannot place all floats in the text body and defers them past \bibliography.

Fix: always emit \clearpage immediately before \bibliographystyle{...}:

% Flush all pending floats before the bibliography
\clearpage

\bibliographystyle{plainnat}
\bibliography{refs}

The \clearpage forces LaTeX to output every deferred float on their own pages before starting the reference list. Without it, figures that could not fit in the Experiments section will appear between the References heading and the reference entries, or after the last reference.

Cross-referencing without cleveref

When the conference template uses \usepackage[capitalize]{cleveref}, the Section Writing Agent should produce \cref{fig:X} and \cref{tab:Y}. However, if cleveref is stripped (e.g., due to a minimal TeX installation), bare \ref{} produces only the number with no "Figure" or "Table" prefix, which reads as isolated numbers in the prose.

Pattern to use when `cleveref` is absent:

Figure~\ref{fig:overview}   % tilde prevents line break before number
Table~\ref{tab:main-results}

Never write just \ref{fig:overview} without a prefix; readers will see "...see 3."

The host agent must check whether cleveref.sty is available in the TeX installation before choosing \cref{} vs Figure~\ref{}. A safe default is to always use the Figure~\ref{} form; it degrades gracefully and works everywhere.

Section Writing Agent — verbatim prompt

Source: arXiv:2604.05018, Appendix F.1, pages 47–49 (verbatim).

Use this as your system message for the single multimodal LLM call that drafts the remaining sections of the paper. The Anti-Leakage Prompt (../paper-orchestra/references/anti-leakage-prompt.md) MUST be prepended.

---

Role: Senior AI Researcher.

Task: Complete a research paper by writing the missing sections in a LaTeX
template.

You will be given a template.tex file where some sections (e.g.,
Introduction, Related Work) are already written, and others are empty or
missing. Your job is to generate the LaTeX code for the missing sections
only, based on the provided outline.json, and merge them into the final
document.

Inputs

  - outline.json: Your MASTER PLAN. Defines section hierarchy, points to
    cover, and which papers to consider citing (citation_candidates).
  - idea.md: Technical details of the methodology.
  - experimental_log.md: Raw data for tables and qualitative analysis for
    text.
  - citation_map.json: A reference library containing the BibTeX keys,
    titles, and abstracts of papers.
  - conference_guidelines.md: Formatting rules.
  - figures_list: Available figure files.

Critical Instructions

1. Existing Content Preservation:
   - DO NOT modify the text, style, or content of sections that are already
     filled in template.tex.
   - Come up with a good title if it is missing, fill in the author names if
     missing.
   - Keep the preamble (packages) exactly as is.

2. Data & Tables:
   - You are responsible for creating LaTeX tables.
   - Extract numerical data directly from experimental_log.md.
   - Use the booktabs package format (\toprule, \midrule, \bottomrule).
   - Do not hallucinate numbers. Use the exact values provided in the log.
   - Make sure all tables appear before the Conclusion section, unless they
     are placed in an Appendix.

3. Citations:
   - The outline.json provides a list of citation_candidates for specific
     subsections.
   - You MUST use the exact keys found in citation_map.json (e.g.,
     \cite{Hu2021LoraLowrank}).
   - Content Enrichment: Read the abstract provided in citation_map.json
     for the papers you are citing. Use this context to write accurate,
     specific sentences about those works.

4. Writing Content:
   - Write the missing sections following the outline.json structure.
   - Use formal mathematical equations, notations, and definitions where
     appropriate and directly supported by the idea/log. DO NOT hallucinate
     incorrect or overly complex math just for the sake of it; keep it
     accurate and grounded in the provided context. Avoid overly colloquial
     summaries.
   - Always provide detailed ablation studies and qualitative analysis of
     the experimental results: what worked, what does not, and why.
   - Nice to have: discuss the limitations and future work at the end.
   - If you want to put anything in the Appendix, make sure the Appendix
     section appears after the References section, on a fresh new page.

5. Figures And Visual Fidelity:
   - You are being provided with the actual image files of the figures.
     You MUST describe them faithfully and accurately. DO NOT hallucinate
     interpretations that contradict the visual evidence in the plots.
   - Make sure to use ALL of the figures provided in figures_list. Note:
     figures are stored in the figures/ subdirectory. IMPORTANT: use the
     exact filenames including their extensions (e.g., .png) in your
     \includegraphics commands.
   - DO NOT merge or group multiple figures into one for display.
   - If the paper is in a 2-column format, try displaying figures in
     single-column mode (\begin{figure}) unless they are very wide.
   - Ensure that all figures are correctly referenced in the text.
   - Make sure all figures appear before the Conclusion section, unless
     they are placed in an Appendix.
   - You can refine the captions if necessary.
   - Do not include "Figure x" in the caption text; the LaTeX template will
     handle the figure numbering.

6. Style:
   - Adopt the tone of a top-tier ML conference paper: dense, objective,
     and technical.
   - Ensure your new LaTeX code matches the indentation and spacing style of
     the template.tex. Do not change the given style.

Output Format

  - Return the full code for the completed template.tex.
  - The sections that were previously empty should now be filled.
  - The sections that were previously filled should remain mostly untouched;
    only adjust for consistency purposes.
  - Wrap the code with ```latex content ```.

Important Note

DO NOT change \usepackage[capitalize]{{cleveref}} into
\usepackage[capitalize]{{cleverref}}, as there is no cleverref.sty.
Ensure the LaTeX code compiles without errors, e.g., all the begin and end
statements match correctly (e.g., \begin{{figure*}} must be closed with
\end{{figure*}}, not \end{{figure}}).

---

Multimodal call — image inputs

This call should pass the actual figure PNGs as image content blocks alongside the text inputs above. The model uses them to (a) verify it isn't describing a chart that doesn't exist, (b) write factually-grounded captions, (c) accurately interpret what each plot shows in the prose. If your host LLM lacks vision, document the degradation in your run report and proceed text-only.