Publishing: Venues, Preprints & Presenting
Conferences, journals and workshops, how to choose a venue, rankings and what scholarship panels count, preprints on arXiv, the camera-ready, and presenting posters and talks.
Where AI research is published
| Type | Examples | Notes |
|---|---|---|
| Top general ML conferences | NeurIPS, ICML, ICLR | highly competitive; 8–9 pages + appendix; deadlines roughly yearly |
| Area conferences | ACL, EMNLP, NAACL (NLP); CVPR, ICCV, ECCV (vision); AAAI, IJCAI; KDD; MICCAI (medical imaging); CHI (HCI) | the main venues for each community; NLP uses a rolling review system (ACL Rolling Review) |
| Workshops | co-located with the conferences above | 4–8 pages, focused topics, often non-archival; excellent for a first paper and feedback |
| Journals | JMLR, TMLR, TPAMI, Nature Machine Intelligence, domain journals | no fixed deadlines; longer review; TMLR focuses on correctness over novelty hype |
| Student and regional venues | student research workshops (e.g. at ACL), regional conferences, AI for social good tracks | mentoring, lower barrier, good first steps |
Many universities and scholarship panels use ranking lists such as the CORE conference rankings (A*, A, B, C) and journal quartiles when counting outputs. Check what your target university counts, but choose venues primarily by fit with the audience who should read your work.
Try it: choose a venue
Describe your paper (maturity, area, timing) and see which kinds of venue fit, and why.
Interactive: this part runs in your browser.
Preprints and the camera-ready
Posting on arXiv makes work public immediately, establishes priority and lets supervisors read it when you apply. Check each venue's policy on preprints during anonymous review (most AI venues allow arXiv, some restrict publicity during review). Use a clear title and abstract; update versions rather than posting new ones.
After acceptance, the camera-ready version de-anonymises the paper, adds acknowledgements (funding, compute, people), includes the promised changes, and the code link. Then register and present.
Presenting: posters and talks
- Poster: title as the claim, one key figure in the centre, three or four boxes (problem, approach, results, takeaway), large fonts, little text, a QR code to the paper and code. Prepare a 30-second and a 3-minute explanation.
- Talk (often 5–15 minutes): one main message; motivate with an example; show one or two key results; end with the takeaway and limitations. One idea per slide; rehearse with a timer.
- At conferences, talk to people at posters, ask questions, and follow up by email: many PhD offers and collaborations start this way.
Quiz
1. For a first, focused paper that needs feedback, a good target is often…
2. CORE rankings are used by many universities to…
3. Why post a preprint before applying for a PhD?
4. A good poster title is…
📚 Further reading
Trusted, mostly free sources: official documentation, well-known textbooks and courses. When you want to confirm or go deeper on anything here, start with these.
- CORE Conference Rankings ↗ rankings used by many Australian panels
- AI Conference Deadlines ↗ community-maintained deadline tracker
- arXiv ↗ preprint server
- S. Peyton Jones, How to Give a Great Research Talk ↗ talk and slides