Skip to content

Sydney 6–12 Dec · Atlanta and Paris 9–13 DecUnofficial guide by QRS

NeurIPS 2026 Attendee Guide

NeurIPS 2026 · Presenting

Presenting an oral

Oral sessions in each city, the room set-up, slide upload, SlidesLive recording, free 16:9 templates and a prompt for an AI assistant.

Summary

Slides are 16:9 and projectors take HDMI; bring your own adapter and laser pointer. Upload a PDF of at most 10 MB at neurips.cc/SlidesUpload and start with a title slide or a QR code to your paper page (general guidance from the conference platform wiki, not year-specific; unconfirmed for 2026). NeurIPS has not published 2026 talk lengths; in Atlanta four papers share each 60-minute session.

Checked against official pages on Thu 5 Nov 2026.

The rules at a glance

Oral sessions in each city

From the official schedule, in each venue's local time. Every oral paper also has a poster slot. NeurIPS has not published how long each talk is.

Sydney 17 sessions · 92 oral papers · AEDT

DayTimeSessions (room)
Tue 8 Dec15:30–16:30 60 min1A, 1B, 1C, 1D, 1E
Wed 9 Dec15:30–16:30 60 min2A, 2B, 2C, 2D, 2E, 2F
Thu 10 Dec15:30–16:30 60 min3A, 3B, 3C, 3D, 3E, 3F

The schedule does not yet list which papers are in each session.

Atlanta 6 sessions · 24 oral papers · EST

DayTimeSessions (room)Papers
Wed 9 Dec15:30–16:30 60 min1A Reliable & Safe LLM Agents and Reasoning (Hall C3), 1B Theory: Online Algorithms, Dynamics & Complexity (C111-C112)4 each
Thu 10 Dec15:30–16:30 60 min2A Generative Modeling: Diffusion, Visual & 4D Generation (Hall C3), 2B Trustworthy Vision: Explanation, Forensics & Identity (C111-C112)4 each
Fri 11 Dec15:30–16:30 60 min3A World Models, 3D Scenes & Collaborative Agents (Hall C3), 3B Benchmarks & Evaluation Methodology (C111-C112)4 each

Paris 10 sessions · 16 oral papers · CET

DayTimeSessions (room)
Wed 9 Dec09:15–09:45 30 min1
Wed 9 Dec09:45–10:15 30 min2
Thu 10 Dec10:00–10:30 30 min3 (Grand Amphithéâtre)
Thu 10 Dec10:30–11:00 30 min4
Fri 11 Dec10:00–10:30 30 min5
Fri 11 Dec10:30–11:00 30 min6
Fri 11 Dec13:30–15:00 90 min9
Fri 11 Dec14:00–14:30 30 min7 (Grand Amphithéâtre)
Fri 11 Dec14:30–15:00 30 min8
Fri 11 Dec16:30–18:00 90 min10

Some sessions overlap: they run at the same time in different rooms. The schedule does not yet list which papers are in each session.

From the official schedule. No room means none is published yet.

Templates and skills

From NeurIPS

No slide template is on the public NeurIPS pages for 2026 as of 5 November 2026, or, as far as we found, for earlier years.

Made by QRS for 2026

3 items

Unofficial, 16:9, MIT licence.

Made by others

2 items

Free beamer themes and galleries; set the aspect ratio to 16:9.

  • Metropolis beamer themeA plain, widely used beamer theme.LaTeX
    Made by othersOpen
  • Overleaf presentation galleryFree beamer templates.LaTeX
    Made by othersOpen

Skills for AI agents (SKILL.md)

7 items

Folders an agent such as Claude Code reads before it works. Ours carries the 2026 rules; the others teach general slide making. Some include scripts; QRS has not reviewed them, so read before running.

  • NeurIPS 2026 talk skill (SKILL.md)The 2026 slide and oral rules, our beamer and PowerPoint templates and a timed-script workflow.SKILL.md
    Made by QRSDownload zip, 3 KBDownload SKILL.md
  • beamer skillCreates, compiles and polishes academic beamer decks, including from a paper PDF. MIT. Copied here unchanged with its licence and source; all credit to its authors. Some include scripts; QRS has not reviewed them, so read before running.SKILL.mdLaTeX
    Made by othersDownload zip, 42 KBDownload SKILL.mdUpstream on GitHub
  • scientific-slides skill (K-Dense)Research talks in PowerPoint or beamer, with structure and timing. MIT. Its image generation needs an OpenRouter key. Copied here unchanged with its licence and source; all credit to its authors. Some include scripts; QRS has not reviewed them, so read before running.SKILL.mdPowerPointLaTeX
    Made by othersDownload zip, 140 KBDownload SKILL.mdUpstream on GitHub
  • marp-slides skill (marpsmith)Slides written in Markdown with Marp, exported to HTML, PDF or PowerPoint. MIT. Copied here unchanged with its licence and source; all credit to its authors. Some include scripts; QRS has not reviewed them, so read before running.SKILL.md
    Made by othersDownload zip, 89 KBDownload SKILL.mdUpstream on GitHub
  • pptx skill (Anthropic)Official Anthropic skill for creating and editing PowerPoint decks. Source-available, all rights reserved by Anthropic.SKILL.mdPowerPoint
    Made by othersOpen
  • pdf skill (Anthropic)Official Anthropic skill for creating, merging and checking PDFs. Source-available, all rights reserved by Anthropic.SKILL.mdPDF
    Made by othersOpen
  • web-artifacts-builder skill (Anthropic)Official Anthropic skill for richer HTML artifacts (React, Tailwind). Apache 2.0. Copied here unchanged with its licence and source; all credit to its authors. Some include scripts; QRS has not reviewed them, so read before running.SKILL.md
    Made by othersDownload zip, 30 KBDownload SKILL.mdUpstream on GitHub

Make it with an AI assistant

Choose the output: beamer and PDF, PowerPoint, a Claude Artifact slide deck, an HTML file, or several at once. Tick the last box for a version that first reads every official page, guide and skill. Paste it into ChatGPT, Claude, Gemini or a coding agent with your paper, fill in the brackets, and check every number against your paper.

Prompt: make my talk slidesPlain text (default)
You are helping me make the slides for my NeurIPS 2026 oral talk. Work only from my paper.

My paper: [attach the PDF, or paste its arXiv or OpenReview link]
Talk length: [minutes. NeurIPS has not published 2026 talk lengths: check your session on neurips.cc or ask the session chair]
Session line: [for example: Atlanta Orals 1A, Wed 9 Dec, 15:30, Hall C3]
Paper page: [https://neurips.cc/virtual/2026/poster/NNNNNN, from neurips.cc/MyStuff]

Rules from NeurIPS and its conference platform (the wiki.eventhosts.cc ones are general platform guidance, not year-specific, and unconfirmed for 2026):
- Slides are 16:9. Projectors take HDMI; I bring my own adapter and laser pointer (neurips.cc/FAQ/AudioVisualProjectorConnection).
- The first slide names the paper and the authors, or has a QR code to the paper page: SlidesLive uses the opening slide to match the recording to the paper (wiki.eventhosts.cc/en/reference/physical-presenter-instructions).
- Upload a PDF of at most 10 MB at neurips.cc/SlidesUpload (general platform guidance; unconfirmed for 2026). Avoid videos embedded in the PDF unless I present from my own laptop. Do not depend on the venue internet.
- Accessibility (neurips.cc/Conferences/2026/MainTrackHandbook): minimal text, large fonts, bold for emphasis, no text effects such as shadows, high contrast (dark text on cream works best), colour-blind safe palettes, no flashing, and nothing important that is only on the slide and not said aloud.

Do this:
1. Read the whole paper. Propose an outline at about one slide a minute: title, the problem, why it is hard, our idea, method, main result, what else we found, takeaways, closing. Give each slide one message as its title and a time in seconds. Show me the outline and wait for my OK.
2. Write the slides. At most about six short bullets on a slide; prefer one figure and one sentence.
3. Use the paper's own figures. For each figure, write the sentence I will say to describe it: the axes first, then what to look at.
4. Never invent numbers, results or citations. Every number must appear in the paper. Where something is missing, write TODO.
5. Write a speaker script, slide by slide, that fits the talk length with a minute to spare.
6. Prepare three likely questions with short answers, taken from the paper's limitations and experiments.

Make the slides in each output below. Make each output from the same content.

Output: LaTeX and PDF
Use the QRS beamer theme (unofficial, made by the Quantitative Research Society, MIT licence). Download:
- https://qrsntu.org/service/neurips26/assets/talk/beamerthemeneurips26.sty
- https://qrsntu.org/service/neurips26/assets/talk/talk.tex
Its commands: \paperurl (the QR code), \paperid, \session, \contact, \placeholder{height}{label}, \takeaway{sentence}, \closingslide{sentence}.
Write talk.tex. Compile with pdflatex twice. Check: 16:9 pages, no overfull boxes, every font readable from the back of a hall, PDF under 10 MB.
If you cannot run LaTeX, give me the files to upload to Overleaf.

Give me back: the files for each of the outputs above, the script, the questions, and a list of every TODO.

What the handbook asks of talks

  1. Handbook: slide text Make slides with minimal text, large fonts and bold for emphasis, and avoid special text effects such as shadows. Main Track Handbook
  2. Handbook: contrast Choose high-contrast colours; the handbook says dark text on a cream background works best. Main Track Handbook
  3. Handbook: flashing and colour Avoid flashing text or graphics. Choose palettes that work for colour-blind viewers and do not rely on colour to convey a message. Main Track Handbook
  4. Handbook: inclusive examples Use examples understandable to a diverse, multicultural audience, and avoid sexual, racial or otherwise offensive text, examples or imagery; use only material you are fully confident is appropriate. Main Track Handbook
  5. Handbook: speaking Do not assume all audience members can see the slides; say aloud everything important. The handbook says its paper accessibility guidance also applies to posters. Main Track Handbook
  6. Guides it points to The handbook recommends three guides for accessible talks: the W3C guide to accessible presentations, the ACM guide, and the RECSYS accessible presentations guide. Main Track Handbook
  7. W3C: describe visuals, limit text Cover all information on each slide aloud without reading word for word, describe visual cues in the room, limit text per slide, put detail in a handout, and offer slides in editable formats rather than only PDF. Use dark text on light in a lit room. W3C WAI
  8. RECSYS guide: concrete numbers Start around 22 pt sans-serif, no more than about six bullets per slide, avoid fully justified text and all caps, add alt text, use a green rather than red laser pointer, and share slides before the talk. RECSYS guide

Your SlidesLive video

  1. Where the link is Recording links appear under Presentation History on your MyStuff page (sign-in needed). Poster Information
  2. Recording Tips: setup Use Google Chrome, a quiet bright room with a clean background, an external microphone if possible (headset mics work), and record in daytime with lights on. SlidesLive tips
  3. Recording Tips: camera Sit straight with the camera at eye level, centred, chest-and-head framing. You may also record on a DSLR or phone and attach the file with the Add File button after recording. SlidesLive tips
  4. Recording Tips: slides Put slides on full screen, hide the mouse cursor and the screen-sharing bar, do not move or shake the laptop or table, introduce yourself and smile. SlidesLive tips
  5. What the tips do not say The Recording Tips page states no video format, length limit or resolution. Check the recording page on MyStuff for those. SlidesLive tips

Guides to read

Still to come

Not published yet

  • Talk lengths for 2026 orals
  • Oral rooms in Sydney and most of Paris
  • A speaker ready room, or whether a presentation laptop is provided
  • The SlidesLive video format and length limit

More in this guide