$npx -y skills add AlexAI-MCP/hermes-CCC --skill research-paper-writingAcademic paper writing workflow — literature review, LaTeX setup, structure, citations, and conference submissions.
| 1 | # Research Paper Writing |
| 2 | |
| 3 | End-to-end workflow for writing and submitting academic papers. |
| 4 | |
| 5 | --- |
| 6 | |
| 7 | ## Paper Structure |
| 8 | |
| 9 | ``` |
| 10 | 1. Title & Authors |
| 11 | 2. Abstract ← write last, 150-250 words |
| 12 | 3. Introduction ← problem, gap, contributions, roadmap |
| 13 | 4. Related Work ← what exists, why insufficient |
| 14 | 5. Methodology ← your approach, architecture, algorithm |
| 15 | 6. Experiments ← setup, baselines, metrics |
| 16 | 7. Results ← tables, figures, analysis |
| 17 | 8. Discussion ← limitations, broader impact |
| 18 | 9. Conclusion ← summary, future work |
| 19 | 10. References |
| 20 | Appendix (optional) |
| 21 | ``` |
| 22 | |
| 23 | --- |
| 24 | |
| 25 | ## Abstract Formula |
| 26 | |
| 27 | ``` |
| 28 | [PROBLEM] Prior work has struggled with X. |
| 29 | [GAP] Existing approaches fail because Y. |
| 30 | [SOLUTION] We propose Z, which does A and B. |
| 31 | [RESULTS] On benchmark C, we achieve D% improvement. |
| 32 | [IMPACT] This enables E. |
| 33 | ``` |
| 34 | |
| 35 | --- |
| 36 | |
| 37 | ## LaTeX Setup |
| 38 | |
| 39 | ```bash |
| 40 | # Ubuntu/Debian |
| 41 | sudo apt install texlive-full |
| 42 | |
| 43 | # macOS |
| 44 | brew install --cask mactex |
| 45 | |
| 46 | # Windows |
| 47 | # Install MiKTeX from miktex.org |
| 48 | |
| 49 | # Verify |
| 50 | pdflatex --version |
| 51 | bibtex --version |
| 52 | ``` |
| 53 | |
| 54 | ### Compile |
| 55 | |
| 56 | ```bash |
| 57 | pdflatex paper.tex |
| 58 | bibtex paper |
| 59 | pdflatex paper.tex |
| 60 | pdflatex paper.tex # run 3x for refs to resolve |
| 61 | |
| 62 | # Or with latexmk (auto-detects what to run) |
| 63 | latexmk -pdf paper.tex |
| 64 | latexmk -pdf -pvc paper.tex # continuous preview |
| 65 | ``` |
| 66 | |
| 67 | --- |
| 68 | |
| 69 | ## Minimal LaTeX Paper |
| 70 | |
| 71 | ```latex |
| 72 | \documentclass{article} |
| 73 | \usepackage{amsmath, amssymb, graphicx, booktabs, hyperref} |
| 74 | \usepackage[margin=1in]{geometry} |
| 75 | \usepackage[numbers]{natbib} |
| 76 | |
| 77 | \title{Your Paper Title} |
| 78 | \author{Author One \and Author Two} |
| 79 | \date{} |
| 80 | |
| 81 | \begin{document} |
| 82 | \maketitle |
| 83 | |
| 84 | \begin{abstract} |
| 85 | Your abstract here. |
| 86 | \end{abstract} |
| 87 | |
| 88 | \section{Introduction} |
| 89 | \label{sec:intro} |
| 90 | |
| 91 | \section{Related Work} |
| 92 | Prior work \citep{vaswani2017attention} showed... |
| 93 | |
| 94 | \section{Method} |
| 95 | \label{sec:method} |
| 96 | |
| 97 | \section{Experiments} |
| 98 | |
| 99 | \begin{table}[t] |
| 100 | \centering |
| 101 | \caption{Main results} |
| 102 | \begin{tabular}{lcc} |
| 103 | \toprule |
| 104 | Method & Metric A & Metric B \\ |
| 105 | \midrule |
| 106 | Baseline & 70.1 & 65.3 \\ |
| 107 | Ours & \textbf{74.8} & \textbf{70.2} \\ |
| 108 | \bottomrule |
| 109 | \end{tabular} |
| 110 | \end{table} |
| 111 | |
| 112 | \section{Conclusion} |
| 113 | |
| 114 | \bibliography{references} |
| 115 | \bibliographystyle{plainnat} |
| 116 | \end{document} |
| 117 | ``` |
| 118 | |
| 119 | --- |
| 120 | |
| 121 | ## References (.bib file) |
| 122 | |
| 123 | ```bibtex |
| 124 | @article{vaswani2017attention, |
| 125 | title={Attention is all you need}, |
| 126 | author={Vaswani, Ashish and others}, |
| 127 | journal={NeurIPS}, |
| 128 | year={2017} |
| 129 | } |
| 130 | |
| 131 | @inproceedings{devlin2019bert, |
| 132 | title={{BERT}: Pre-training of deep bidirectional transformers}, |
| 133 | author={Devlin, Jacob and others}, |
| 134 | booktitle={NAACL}, |
| 135 | year={2019} |
| 136 | } |
| 137 | ``` |
| 138 | |
| 139 | Get BibTeX from: |
| 140 | - Google Scholar → Cite → BibTeX |
| 141 | - arXiv → Export BibTeX (bottom of abstract page) |
| 142 | - Semantic Scholar |
| 143 | |
| 144 | --- |
| 145 | |
| 146 | ## Literature Review via arXiv |
| 147 | |
| 148 | ```bash |
| 149 | # Use /arxiv skill to find papers |
| 150 | # Example searches: |
| 151 | # /arxiv search "GRPO reinforcement learning language model" |
| 152 | # /arxiv search "retrieval augmented generation survey" |
| 153 | # /arxiv get 2401.00001 |
| 154 | ``` |
| 155 | |
| 156 | --- |
| 157 | |
| 158 | ## Figures |
| 159 | |
| 160 | ```python |
| 161 | import matplotlib.pyplot as plt |
| 162 | import numpy as np |
| 163 | |
| 164 | fig, ax = plt.subplots(figsize=(4, 3)) |
| 165 | x = np.linspace(0, 10, 100) |
| 166 | ax.plot(x, np.sin(x), label="Our Method") |
| 167 | ax.plot(x, np.cos(x), "--", label="Baseline") |
| 168 | ax.set_xlabel("Epoch") |
| 169 | ax.set_ylabel("Accuracy") |
| 170 | ax.legend() |
| 171 | ax.grid(alpha=0.3) |
| 172 | plt.tight_layout() |
| 173 | plt.savefig("figure1.pdf", bbox_inches="tight", dpi=300) |
| 174 | ``` |
| 175 | |
| 176 | Include in LaTeX: |
| 177 | ```latex |
| 178 | \begin{figure}[t] |
| 179 | \centering |
| 180 | \includegraphics[width=0.8\linewidth]{figure1.pdf} |
| 181 | \caption{Comparison of our method vs baseline.} |
| 182 | \label{fig:main} |
| 183 | \end{figure} |
| 184 | ``` |
| 185 | |
| 186 | --- |
| 187 | |
| 188 | ## Conference Templates |
| 189 | |
| 190 | | Conference | Deadline | Template | |
| 191 | |-----------|---------|---------| |
| 192 | | NeurIPS | May | neurips.cc | |
| 193 | | ICLR | Oct | iclr.cc | |
| 194 | | ICML | Jan | icml.cc | |
| 195 | | ACL | Feb | acl-org.github.io | |
| 196 | | AAAI | Aug | aaai.org | |
| 197 | |
| 198 | Download official template from the conference website — never use unofficial ones. |
| 199 | |
| 200 | --- |
| 201 | |
| 202 | ## Overleaf (Collaborative) |
| 203 | |
| 204 | 1. Upload all `.tex`, `.bib`, `.pdf` files to Overleaf |
| 205 | 2. Share link with co-authors |
| 206 | 3. Enable Track Changes for reviews |
| 207 | 4. Download final PDF when ready |
| 208 | |
| 209 | --- |
| 210 | |
| 211 | ## Common Mistakes |
| 212 | |
| 213 | - **Vague claims**: "significantly better" → give exact numbers |
| 214 | - **Missing baselines**: always compare against SOTA |
| 215 | - **No ablation**: show which parts of your method matter |
| 216 | - **Figure too small**: aim for readable at 100% zoom |
| 217 | - **Passive voice overuse**: "we propose" not "it is proposed" |
| 218 | - **Missing limitations**: reviewers will point them out; address proactively |