Project 14 — Final Paper, Symposium Talk, and Publication Roadmap (FINAL)

“Whatever you do, work heartily, as for the Lord and not for men, knowing that from the Lord you will receive the inheritance as your reward.” — Colossians 3:23–24 (ESV)

Chapter: 16 — The Research Symposium Due: end of Week 16 (symposium day for the talk; paper + roadmap submitted with it) Submit: to Canvas — (1) the finished paper as PDF, (2) the slide deck, (3) roadmap.txt, (4) agent-log.txt, and (5) a link to your tagged portfolio repo Allowed tools: open everything — your AI assistant, Overleaf, Zotero, Python, the literature tools. This is the work, not a closed exam. AI policy: open-AI for the work, with disclosure and integrity per Appendix C. An agent-log.txt is required. The model is never an author and never an excuse — you own every word, figure, citation, and line of code. Weight: Final Paper 15% + Final Presentation 5% of the course grade. Graded most heavily on the paper and the roadmap — the documents that carry your judgment.

The Setup

It is the last week. There is a room: your cohort, your instructor, maybe a faculty guest or two. You have twelve minutes and a finished paper. Beyond the room is something larger — a real venue with a real call for papers and a real deadline, and the question this whole course was built to make you answer: is this real research, or was it a class project wearing research clothes?

You answer it three ways this week. You stand up and give the talk — the trailer for the paper, built to leave the room able to repeat one true sentence about what you found. You hand in the finished manuscript, formatted for an actual venue, every citation verified, every claim matched to the data. And you write the roadmap: an honest assessment of the gap between what you have and what a venue requires, and a concrete plan to close it. The talk is performance. The paper is the artifact. The roadmap is the judgment — and it is the part no model can write for you.

This is the first half of a thesis, finished. Treat it like one.

Learning Targets

You will demonstrate that you can:

• Compress a research paper into a 12-minute conference-style talk built around one remembered claim.
• Field questions honestly — answering what you can, saying “I didn’t test that” when you can’t, never bluffing.
• Finish a paper in a real CS venue template (acmart sigconf or IEEEtran conference), in IMRaD-for-CS structure.
• Verify every citation against a real index and disclose AI use per a venue’s policy.
• Choose a realistic venue from its CFP, CORE rank, and h5-index, and map the path from draft to submission.
• Make and defend a judgment a tool cannot make: the honest gap assessment and the venue choice.

Normal Tier

Meets the deliverable bar for the final: a complete talk, a finished formatted paper, and a real roadmap.

Required features

1. The talk. A 12-minute (±1 min) symposium presentation, ~10–12 slides, one beat per slide per code/talk-outline.txt, built around one explicit remembered sentence. Delivered live; 3 minutes of Q&A.
2. The paper. The finished manuscript in acmart sigconf or IEEEtran conference format, compiling clean on Overleaf, in IMRaD-for-CS structure (abstract / intro / related work / method / results / discussion / future work), fitting a real venue’s page limit in that venue’s count format.
3. Honest results. The headline result reported with at least the limitations stated plainly; no single-run claim presented as if it were robust without saying so.
4. Verified citations. Every reference verified against dblp / Semantic Scholar / the publisher via code/citation_audit.py; the filled citation_audit.csv included. Zero unverified references.
5. AI disclosure. A disclosure paragraph in the location your chosen venue requires (§16.7); no hidden prompt text anywhere.
6. The roadmap. A completed roadmap.txt from code/submission-roadmap.txt naming one real venue (read its actual CFP), with code/venue-fit-sheet.txt filled in, real deadline dates, and a continuation plan.
7. agent-log.txt. A running log of substantive AI use this week.

Normal-tier rubric (out of 100)

Medium Tier (+up to 25% extra credit)

Stronger rigor and a more credible submission.

• Multi-seed headline result with effect size and a confidence interval, not a single run (Bouthillier et al. 2021), reported correctly in the results section.
• A reproducibility artifact: environment pinned, seeds set, data documented, a tagged Git release, and a Zenodo deposit with a DOI (auto-linked Software Heritage SWHID).
• A polished, anonymized submission build ready for double-anonymous review (names/affiliations/self-citations/grant IDs scrubbed).
• Three rehearsed backup slides that visibly answer the audience’s hardest questions during Q&A.

Hard Tier (+up to 25% additional extra credit)

Publication-ready, and a decision an AI cannot make for you.

Write a 2–3 page venue-decision memo that does what no tool can: it commits. Compare at least three real candidate venues on fit, selectivity (CORE rank, h5-index, illustrative acceptance rate cited to a source and year), deadline feasibility for your current results, archival-vs-non-archival tradeoff, and AI-disclosure burden — and then recommend one, and defend the recommendation against the strongest case for a different choice. The memo must name what would have to be true about your results to justify aiming one tier higher, and what your continuation plan is if the first submission is rejected. The grade is on the quality of the judgment, not the prestige of the venue you pick.

Submission

Submit to Canvas:

1. paper.pdf — the finished, venue-formatted manuscript.
2. slides.pdf (or a link) — the symposium deck.
3. roadmap.txt — the completed publication roadmap, with the venue-fit sheet.
4. citation_audit.csv — every reference verified.
5. agent-log.txt — your AI-use log for the week.
6. A link to your tagged portfolio repo (git tag for the symposium release).
7. (Medium/Hard) the Zenodo DOI and/or the venue-decision memo.

Deliver the talk live at the symposium.

Hints

• Write the one remembered sentence first (Rep 1). Build everything toward it.
• Do not start a new experiment this week. Reformat, tighten, and finish. New experiments go in the continuation plan.
• Run code/citation_audit.py early — finding a hallucinated citation the night before is avoidable panic.
• Pick the venue before you finish formatting, so you format into the right template and page limit the first time.
• Rehearse the talk out loud, twice, on a clock. The first run always overruns.
• When in doubt about disclosure, disclose — in the Acknowledgements, confirmed against the CFP.

What Mastery Looks Like

A mastered final is one a stranger could pick up cold and believe: a talk that leaves the room repeating one true sentence; a paper whose every citation resolves and whose every claim is matched to its evidence; a results section that reports effect size and a confidence interval, not a lone lucky run; a disclosure paragraph in exactly the right place; and a roadmap that names a real venue, a real deadline, and the honest gap to close — a plan you could actually execute next month. Mastery is not the most impressive paper in the room. It is the most trustworthy one.

Coach’s Note — The strongest finals I have graded were not the flashiest results. They were the ones where the student told me the truth about their data, picked a venue they could actually reach, and handed me a plan I believed they would follow. That student is a researcher. The flashy-result student who overclaimed is, for now, a student who got a result. Be the first one.

When You’re Done

• One remembered sentence written; talk built around it.
• Talk rehearsed out loud, twice, lands inside 12 minutes.
• Three backup slides for the hardest questions.
• Paper in the venue template, compiles, fits the page limit, IMRaD-for-CS.
• Results honest; limitations stated; no overclaiming.
• Every citation verified; citation_audit.csv complete.
• AI disclosure in the venue’s required location; no hidden text.
• roadmap.txt names a real venue, real dates, a gap assessment, a continuation plan.
• agent-log.txt honest and complete; repo tagged.
• You can stand behind every word in front of a room.

A theological footnote. The verse is Colossians 3:23–24 — whatever you do, work heartily, as for the Lord and not for men. The symposium is the moment the pull to work for men is strongest: the room, the grade, the line on a CV. That motive cannot survive a research career — it will be rejected, it will find its own bug, it will read an unfair review — and as it fails it curdles into the exact temptations this course trained you against: the dropped seed, the cherry-picked run, the unverified citation, the slide that claims a little more than the data. There is one motive that holds, because its audience already sees the data you are tempted to hide. To work as for the Lord is not a decoration on good method; it is the root of it — the reason honest statistics, verified citations, and reproducible code are worth the extra hours when no one is checking. Give the best talk you can. Submit the best paper you can. Aim at a real venue and mean it. And do the work for an audience of One, which is also, not by accident, the only way to do research you will still respect in ten years.