Agent Skills

presenting-conference-talks

Generates conference presentation slides (Beamer LaTeX PDF and editable PPTX) from a compiled paper with speaker notes and talk script. Use when preparing oral talks, spotlight presentations, or invited talks for ML and systems conferences.

Install

npx skills add https://github.com/orchestra-research/ai-research-skills --skill presenting-conference-talks
SKILL.md

Presenting Conference Talks: From Paper to Slides

Generate conference presentation slides from a compiled research paper. Produces both Beamer LaTeX PDF (for polished typesetting) and editable PPTX (for last-minute adjustments), with speaker notes and an optional talk script.

When to Use This Skill

Scenario Use This Skill Use Other Skills Instead
Preparing oral/spotlight/poster-talk slides ✅
Generating Beamer PDF + PPTX from paper ✅
Speaker notes and talk script ✅
Writing the paper itself ml-paper-writing
Structuring a systems paper systems-paper-writing
Creating publication-quality plots academic-plotting

Attribution: This skill's structure draws inspiration from the ARIS paper-slides skill (570 lines, supporting poster/spotlight/oral/invited with Beamer+PPTX). This is an independent implementation for the AI-Research-SKILLs ecosystem.


Talk Types and Slide Counts

Talk Type Duration Slides Content Depth
poster-talk 3–5 min 5–8 Problem + key result only
spotlight 5–8 min 8–12 Problem + approach + key results
oral 15–20 min 15–22 Full story with evaluation highlights
invited 30–45 min 25–40 Deep dive with context and demos

Rule of thumb: ~1 slide per minute for oral, ~1.5 slides per minute for spotlight.


Slide Structure Templates

Poster-Talk (5–8 slides)

Slide 1: Title + Authors + Affiliation
Slide 2: Problem — Why this matters (1 motivating figure)
Slide 3: Key Insight — One-sentence thesis
Slide 4: Approach Overview — Architecture diagram
Slide 5: Main Result — Headline numbers (1 figure)
Slide 6: Takeaway + QR code to paper/code

Spotlight (8–12 slides)

Slide 1:  Title + Authors
Slide 2:  Problem Statement — Concrete, quantified
Slide 3:  Motivation — Why existing solutions fall short
Slide 4:  Key Insight — Thesis statement
Slide 5:  System Overview — Architecture diagram
Slide 6:  Design Highlight 1 — Core mechanism
Slide 7:  Design Highlight 2 — Key innovation
Slide 8:  Evaluation Setup — Baselines and workloads (brief)
Slide 9:  Main Results — Headline performance figure
Slide 10: Ablation / Breakdown — What contributes most
Slide 11: Summary + Contributions
Slide 12: Thank You + Links

Oral (15–22 slides)

Slide 1:  Title + Authors + Venue
Slide 2:  Outline (optional — "roadmap" slide)
Slide 3:  Problem Context — Domain importance
Slide 4:  Problem Statement — Specific challenge
Slide 5:  Motivation — Gaps in existing systems
Slide 6:  Key Insight — Thesis
Slide 7:  System Overview — Architecture diagram
Slide 8:  Design Component 1 — Detailed walkthrough
Slide 9:  Design Component 2 — Detailed walkthrough
Slide 10: Design Component 3 — Detailed walkthrough
Slide 11: Design Alternatives — Why not other approaches
Slide 12: Implementation — Key engineering highlights
Slide 13: Evaluation Setup — Testbed, baselines, metrics
Slide 14: End-to-End Results — Main performance
Slide 15: Result Deep Dive — Breakdown or per-workload
Slide 16: Ablation Study — Component contributions
Slide 17: Scalability — Scaling behavior
Slide 18: Demo Slide (systems talks) — Screenshot or recording
Slide 19: Related Work — Positioning (brief)
Slide 20: Summary — Contributions restated
Slide 21: Future Work — Open questions
Slide 22: Thank You + Paper Link + QR Code

Invited Talk (25–40 slides)

Extends the oral structure with:

  • Additional context slides (field overview, historical progression)
  • Multiple demo/walkthrough slides
  • Deeper evaluation analysis
  • Broader implications and future directions
  • Q&A preparation slides (hidden, for backup)

Systems Talk Specifics

Systems conference talks have unique requirements compared to ML talks:

Demo Slide

  • Include a live demo or pre-recorded screencast of the system in action
  • Always have a recorded backup — live demos fail at the worst times
  • Show the system under realistic load, not toy examples

Architecture Walkthrough

  • Animate the architecture diagram: highlight components as you explain them
  • Use Beamer \only<N> or \onslide<N> for progressive reveal
  • Walk through a concrete request end-to-end through the system

Evaluation Highlights

  • Select 2–3 strongest figures from the paper
  • Annotate figures on slides (arrows, circles highlighting key points)
  • State the takeaway before showing the figure ("Our system is 2x faster — here's the data")

Speaker Notes Guidelines

Structure per Slide

[Timing: X minutes]
[Key point to convey]
[Transition sentence to next slide]

Mike Dahlin's Layered Approach

Apply "Say what you're going to say, say it, then say what you said" at three levels:

  1. Talk level: Outline slide → body → summary slide
  2. Section level: Section heading → content slides → section takeaway
  3. Slide level: Headline statement → supporting evidence → transition

Timing Guidelines

  • Poster-talk: 30–60 sec per slide
  • Spotlight: 30–45 sec per slide
  • Oral: 45–90 sec per slide
  • Invited: 60–120 sec per slide

Output Formats

Beamer LaTeX → PDF

Advantages: Professional typesetting, math support, version control friendly.

\documentclass[aspectratio=169]{beamer}
\usetheme{metropolis}  % Clean, modern theme
\usepackage{appendixnumberbeamer}

\title{Your Paper Title}
\subtitle{Venue Year}
\author{Author 1 \and Author 2}
\institute{Institution}
\date{}

\begin{document}
\maketitle

\begin{frame}{Problem}
  \begin{itemize}
    \item Key problem statement
    \item Concrete motivation with numbers
  \end{itemize}
  \note{Speaker note: Start with the big picture...}
\end{frame}

% ... more frames ...
\end{document}

python-pptx → Editable PPTX

Advantages: Easy last-minute edits, corporate template compatibility, animations.

from pptx import Presentation
from pptx.util import Inches, Pt
from pptx.enum.text import PP_ALIGN

prs = Presentation()
prs.slide_width = Inches(13.333)  # 16:9
prs.slide_height = Inches(7.5)

# Title slide
slide = prs.slides.add_slide(prs.slide_layouts[0])
slide.shapes.title.text = "Your Paper Title"
slide.placeholders[1].text = "Author 1, Author 2\nVenue Year"

# Content slide
slide = prs.slides.add_slide(prs.slide_layouts[1])
slide.shapes.title.text = "Problem Statement"
body = slide.placeholders[1]
body.text = "Key point 1\nKey point 2"

# Add speaker notes
notes_slide = slide.notes_slide
notes_slide.notes_text_frame.text = "Speaker note: explain the motivation..."

prs.save("talk.pptx")

Color Scheme Suggestions

These are aesthetic suggestions, not official venue requirements. Adjust freely.

Venue Type Primary Accent Background
USENIX (OSDI/NSDI) Dark Blue (#003366) Red (#CC0000) White
ACM (SOSP/ASPLOS) ACM Blue (#0071BC) Dark Gray (#333333) White
NeurIPS Purple (#7B2D8E) Gold (#F0AD00) White
ICML Teal (#008080) Orange (#FF6600) White
Generic Dark Gray (#333333) Blue (#0066CC) White

Workflow

Step 1: Content Extraction

- Read the compiled paper (PDF or LaTeX source)
- Identify: thesis, contributions, architecture figure, key eval figures
- Note the talk type and duration

Step 2: Outline Generation

- Select the appropriate slide structure template (above)
- Map paper sections to slide groups
- Allocate time per slide group

Step 3: Slide-by-Slide Generation

- Generate Beamer source slide by slide
- Add speaker notes per slide
- Include figures from paper (copy to slides/ directory)
- Generate python-pptx script for PPTX version

Step 4: Review and Polish

- Check total slide count matches talk duration
- Verify all figures are readable at presentation resolution
- Run Beamer compilation: latexmk -pdf slides.tex
- Run PPTX generation: python3 generate_slides.py
- Review speaker notes for timing and transitions

Quick Checklist

  • Slide count appropriate for talk type/duration
  • Title slide has correct authors, affiliations, venue
  • Architecture diagram included and clearly labeled
  • Key eval figures annotated with takeaways
  • Speaker notes include timing markers
  • Transitions between sections are smooth
  • Demo slide has recorded backup
  • Thank-you slide includes paper link / QR code
  • Font sizes ≥ 24pt for readability from back of room
  • Consistent color scheme throughout

Common Issues and Solutions

Issue Solution
Too many slides for time limit Cut details, keep one figure per point
Slides feel like paper paragraphs Use bullet points (≤ 6 per slide), let figures tell the story
Audience lost during design section Add architecture walkthrough with progressive reveal
Evaluation slides overwhelming Show 2–3 strongest figures, put rest in backup slides
Speaker notes too long Target 3–4 sentences per slide, focus on transitions
Beamer compilation fails Check figure paths, use \graphicspath{{figures/}}
PPTX looks different from Beamer Adjust python-pptx font sizes and margins manually

References

Related skills

researchmattpocock575KInvestigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent.paper-context-resolverlllllllama451KRigor Paper Context helper for README-first deep learning repo reproduction. Use only when the README and repository files leave a narrow reproduction-critical gap and the task is to resolve a specific paper detail such as dataset split, preprocessing, evaluation protocol, checkpoint mapping, or runtime assumption from primary paper sources while recording conflicts. Do not use for general paper summary, repo scanning, environment setup, command execution, title-only paper lookup, or replacing Renv-and-assets-bootstraplllllllama450KRigor Setup skill for README-first deep learning repo reproduction. Use when the task is specifically to prepare a conservative conda-first environment, checkpoint and dataset path assumptions, cache location hints, and setup notes before any run on a README-documented repository. Do not use for repo scanning, full orchestration, paper interpretation, final run reporting, or generic environment setup that is not tied to a specific reproduction target.ai-research-explorelllllllama311KRigor Explore compatible skill slug for meaningful and potentially novel deep learning research candidates. Use when the researcher has chosen the task family, dataset, benchmark, evaluation method, provided SOTA references, and wants candidate-only exploration on top of `current_research` with auditable repo understanding, idea gating, fair comparison, and governed experiments written to `explore_outputs/`. Do not use for README-first trusted reproduction, open-ended direction finding, narrow c

Search skills and MCP servers

Fuzzy search across 23,137 skills and servers