Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Templates and formatting guides for major academic conference submissions
.claude/skills/brycewang-stanford-conference-paper-template/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 102% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 108% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 75% | 0% |
| case-02 | ✓→✗ | ▼ Worse | 18% | 0% |
| case-03 | ✓→✓ | = Same ✓ | 10% | 0% |
A skill providing templates, formatting guidelines, and submission checklists for major academic conferences. Covers LaTeX and Word template setup, common formatting requirements, and camera-ready preparation.
| Conference/Style | Page Limit | Columns | Font | Margins | Template | |-----------------|-----------|---------|------|---------|----------| | ACL/EMNLP (ARR) | 8 + refs | 2 | Times 11pt | 1in all | acl.sty | | NeurIPS | 8 + refs | 2 | Times 10pt | 1in top/bottom, 0.75in sides | neurips.sty | | ICML | 8 + refs | 2 | Times 10pt | 1in all | icml.sty | | AAAI | 7 + 1 refs | 2 | Times 10pt | 0.75in all | aaai.sty | | IEEE (CVPR, etc.) | 8 | 2 | Times 10pt | 1in top, 0.75in sides | IEEEtran.cls | | ACM (SIGCHI, etc.) | varies | 1 or 2 | Linux Libertine | varies | acmart.cls | | Springer LNCS | 12-16 | 1 | Computer Modern 10pt | varies | llncs.cls |
latex\documentclass[11pt,a4paper]{article} \usepackage[utf8]{inputenc} \usepackage[T1]{fontenc} \usepackage{times} \usepackage{amsmath,amssymb} \usepackage{graphicx} \usepackage{booktabs} \usepackage{hyperref} \usepackage{natbib} \usepackage[margin=1in]{geometry} % Conference-specific: anonymization for review \usepackage{xspace} \newcommand{\etal}{\textit{et al.}\xspace} \title{Your Paper Title: Subtitle with Key Contribution} % For review submission (anonymous) \author{Anonymous Authors} % For camera-ready (de-anonymized) % \author{First Author\textsuperscript{1} \and % Second Author\textsuperscript{2} \\ % \textsuperscript{1}University A, \textsuperscript{2}University B \\ % \texttt{\{first,second\}@email.edu}} \begin{document} \maketitle \begin{abstract} Your abstract here (typically 150-250 words). State the problem, approach, key results, and significance. \end{abstract} \section{Introduction} % Paragraph 1: Problem and motivation % Paragraph 2: Limitations of existing approaches % Paragraph 3: Your approach and contributions % Paragraph 4: Paper organization (optional) \section{Related Work} % Organized by theme, not chronologically \section{Method} % Detailed enough to reproduce \section{Experiments} \subsection{Experimental Setup} \subsection{Results} \subsection{Analysis} \section{Conclusion} \bibliography{references} \bibliographystyle{plainnat} \end{document}
pythondef camera_ready_checklist(paper_info: dict) -> list[dict]: """ Generate a camera-ready preparation checklist. Args: paper_info: Dict with 'venue', 'page_limit', 'has_appendix', etc. """ checks = [ { 'item': 'De-anonymization', 'description': 'Author names, affiliations, and acknowledgments restored', 'critical': True }, { 'item': 'Page limit compliance', 'description': f"Paper is within {paper_info.get('page_limit', 8)} page limit", 'critical': True }, { 'item': 'Copyright/license form', 'description': 'Signed and uploaded to submission system', 'critical': True }, { 'item': 'Figures at 300 DPI', 'description': 'All figures are high resolution (300 DPI minimum for print)', 'critical': True }, { 'item': 'Fonts embedded', 'description': 'All fonts embedded in PDF (check with pdffonts)', 'critical': True }, { 'item': 'References complete', 'description': 'All references have venue, year, pages; no "to appear" remaining', 'critical': True }, { 'item': 'Hyperlinks work', 'description': 'All URLs and DOIs are valid and clickable', 'critical': False }, { 'item': 'Supplementary materials', 'description': 'Code, data, appendices uploaded separately if required', 'critical': False }, { 'item': 'Metadata in PDF', 'description': 'PDF title and author metadata set correctly', 'critical': False } ] return checks
bash# Check for font embedding issues pdffonts paper.pdf # Embed all fonts (if any are not embedded) gs -dNOPAUSE -dBATCH -sDEVICE=pdfwrite \ -dPDFSETTINGS=/prepress \ -dEmbedAllFonts=true \ -dSubsetFonts=true \ -sOutputFile=paper_embedded.pdf \ paper.pdf # Check page count pdfinfo paper.pdf | grep Pages # Verify PDF/A compliance (some venues require it) verapdf paper.pdf
latex% Use booktabs for professional tables (never use vertical lines) \begin{table}[t] \centering \caption{Comparison of methods on benchmark datasets. Best results in \textbf{bold}, second best \underline{underlined}.} \label{tab:results} \begin{tabular}{lcccc} \toprule Method & Dataset A & Dataset B & Dataset C & Avg. \\ \midrule Baseline & 78.2 & 72.1 & 81.3 & 77.2 \\ Previous SOTA & 82.4 & 76.8 & 84.1 & 81.1 \\ Ours (w/o X) & \underline{84.1} & \underline{78.2} & \underline{85.7} & \underline{82.7} \\ Ours (full) & \textbf{86.3} & \textbf{80.1} & \textbf{87.2} & \textbf{84.5} \\ \bottomrule \end{tabular} \end{table}
latex% Figures should appear at top of page or on their own page \begin{figure}[t] \centering \includegraphics[width=\columnwidth]{figures/architecture.pdf} \caption{Overview of the proposed architecture. The encoder processes input sequences while the decoder generates structured output.} \label{fig:architecture} \end{figure}
latexdiff to generate a diff PDF showing changes from previous versions| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 28,594 | 31,469 | +10% | 1 | 1 | 0% | 5,403 | 4,622 | -14% | 0 | 0 | — |
case-02 | pass→fail | 24,453 | 14,571 | -40% | 1 | 1 | 0% | 3,734 | 4,397 | +18% | 0 | 0 | — |
case-03 | pass→pass | 12,089 | 3,586 | -70% | 1 | 1 | 0% | 2,144 | 2,369 | +10% | 0 | 0 | — |
case-04 | fail→pass | 8,054 | 5,192 | -36% | 1 | 1 | 0% | 1,337 | 2,706 | +102% | 0 | 0 | — |
case-05 | fail→pass | 8,342 | 6,884 | -17% | 1 | 1 | 0% | 1,411 | 2,934 | +108% | 0 | 0 | — |
case-06 | fail→fail | 6,286 | 8,199 | +30% | 1 | 1 | 0% | 1,110 | 2,946 | +165% | 0 | 0 | — |
case-07 | pass→pass | 11,515 | 7,499 | -35% | 1 | 1 | 0% | 1,724 | 2,950 | +71% | 0 | 0 | — |
case-08 | fail→fail | 9,615 | 4,909 | -49% | 1 | 1 | 0% | 1,420 | 2,591 | +82% | 0 | 0 | — |
case-09 | fail→fail | 13,007 | 14,716 | +13% | 1 | 1 | 0% | 2,181 | 3,833 | +76% | 0 | 0 | — |
case-10 | pass→pass | 12,627 | 11,064 | -12% | 1 | 1 | 0% | 2,170 | 3,591 | +65% | 0 | 0 | — |
case-11 | pass→pass | 14,038 | 10,348 | -26% | 1 | 1 | 0% | 1,832 | 3,476 | +90% | 0 | 0 | — |
case-12 | pass→pass | 4,103 | 4,867 | +19% | 1 | 1 | 0% | 703 | 2,331 | +232% | 0 | 0 | — |
case-13 | pass→pass | 12,606 | 2,128 | -83% | 1 | 1 | 0% | 1,871 | 2,022 | +8% | 0 | 0 | — |
case-14 | fail→pass | 12,467 | 12,757 | +2% | 1 | 1 | 0% | 1,790 | 3,136 | +75% | 0 | 0 | — |
case-15 | pass→pass | 13,822 | 2,870 | -79% | 1 | 1 | 0% | 1,364 | 2,207 | +62% | 0 | 0 | — |
case-16 | pass→pass | 9,930 | 3,046 | -69% | 1 | 1 | 0% | 1,346 | 2,311 | +72% | 0 | 0 | — |
case-17 | pass→pass | 16,392 | 9,525 | -42% | 1 | 1 | 0% | 2,161 | 3,327 | +54% | 0 | 0 | — |
case-18 | pass→pass | 16,476 | 2,589 | -84% | 1 | 1 | 0% | 2,313 | 2,232 | -4% | 0 | 0 | — |
case-19 | pass→pass | 15,131 | 8,917 | -41% | 1 | 1 | 0% | 2,031 | 3,149 | +55% | 0 | 0 | — |
case-20 | pass→pass | 21,302 | 19,987 | -6% | 1 | 1 | 0% | 3,050 | 4,471 | +47% | 0 | 0 | — |
case-21 | pass→pass | 17,422 | 18,577 | +7% | 1 | 1 | 0% | 2,506 | 4,501 | +80% | 0 | 0 | — |
case-22 | pass→pass | 19,458 | 15,664 | -19% | 1 | 1 | 0% | 2,876 | 4,477 | +56% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +9 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.