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.