Guide
AISTATS vs UAI: not a choice, a queue
AISTATS 2026 notified on 22 January. UAI 2026 closed on 25 February. A theory paper turned down at the first arrives at the second with three to four full reviews, a meta-review and 34 days to use them.
Published 14 August 2026
Still deciding which of them goes first? Paste the abstract lower on this page and see which rooms publish work like yours. Jump to the matcher
The short version
- Send it to AISTATS first. Its decision lands with the UAI deadline five weeks out, and the reverse is not true.
- Then do not carry the numbers across. The same digit means different things at the two venues.
- At AISTATS the accept line falls between 3 and 5. Borderline Reject and Borderline Accept, with the neutral 4 discouraged.
- At UAI a 4 is Borderline reject. There is no neutral seat on that scale at all.
The window
The two calendars interlock, they do not compete
One at a time they look like alternatives. Side by side, the second is where the first one's reviews go.
| AISTATS 2026 | UAI 2026 | |
|---|---|---|
| Registration step | Abstract 25 September 2025 AOE | Submissions open 25 January 2026 |
| Paper deadline | 2 October 2025 AOE, supplementary 9 October | 25 February 2026 AoE |
| Reviewing | 14 October to 10 November 2025 | No window published. Reviews out 23 April 2026 |
| Author response | Rebuttal from 22 November, discussion 1 to 8 December 2025 | 23 April to 2 May, then chairs and reviewers alone to 8 May |
| Decisions | 22 January 2026 AOE | 1 June 2026 |
| Camera ready | 31 March 2026 | 8 July 2026 |
| Conference | 2 to 5 May 2026, Morocco | 17 to 21 August 2026, Amsterdam |
| Second route in | Journal-to-Conference: submit 31 January, decided 15 February 2026 | Nothing comparable |
From the AISTATS 2026 and UAI 2026 dates pages, the prior edition from AISTATS 2025 and the notification date from NeurIPS 2026, read August 2026, and the AISTATS 2027 key dates read 10 September 2026. AISTATS gives conference weekdays that do not match the 2026 calendar, so only the numeric range is printed here. The 34 days are ours.
Thirty-four days is a repair window, not a panic
You do not start from a blank form. You hold three to four reviews and a meta-review on the version you already have, and UAI scores correctness, novelty, evidence, reproducibility and clarity separately, so the objections arrive pre-sorted into the axes scored next. Five weeks buys evidence and writing. It does not buy a new theorem.
AISTATS 2027 is dated now, and it lands where the pattern said
The key dates page gives 29 September 2026 for the abstract and 6 October for the paper, anywhere on earth, adding that they may be subject to change. The last two editions closed 10 October 2024 and 2 October 2025. NeurIPS 2026 notifies on 24 September 2026, so the gap is five days to an abstract, not the month people plan for.
The two scales
A 4 is a fence-sit at one venue and a rejection at the other
Both publish the whole form with its anchors, which is rare. Lay the two overall scores against one column of numbers and no score survives the trip.
| Number on the form | AISTATS 2026, 7 points | UAI 2026, 10 points |
|---|---|---|
| 10 | No such score | Award quality |
| 9 | No such score | Very Strong Accept |
| 8 | No such score | Strong Accept |
| 7 | Strong Accept. Excellent, clearly novel, impact beyond a single sub-area of AI and stats | Accept |
| 6 | Accept. Solid, high impact on at least one sub-area | Weak Accept |
| 5 | Borderline Accept. Reasons to accept outweigh reasons to reject | Borderline accept |
| 4 | Weak Accept and Weak Reject at once. Merits and weaknesses roughly balanced, and to be used sparingly | Borderline reject |
| 3 | Borderline Reject. Reasons to reject outweigh reasons to accept | Reject |
| 2 | Reject. Technical flaws, weak evaluation, inadequate reproducibility | Strong Reject |
| 1 | Strong Reject. Well-known results, or unaddressed ethical considerations | Very Strong Reject |
Anchors condensed from the AISTATS 2026 reviewer guidelines and the UAI 2026 reviewer and area chair instructions, read August 2026. Both also run a 1 to 5 confidence scale. Read the ends: AISTATS separates 7 from 6 by impact beyond a single sub-area, and offers well-known results as its example of a 1, pricing rediscovery as a strong reject.
The tinted row is the trap
At AISTATS the line runs between 3 and 5, and the 4 is a seat the form discourages: the reviewer declined to choose. At UAI the line runs between 4 and 5, no neutral position exists, and a 4 is a reviewer choosing against you. Reading last season's scores across is how authors talk themselves into the wrong rebuttal.
| Axis | AISTATS, categorical options | UAI, question 2, each 1 to 4 |
|---|---|---|
| Novelty | Ground-breaking, New results, Incremental, Known results | Originality: 4 ground-breaking, 3 non-trivial advance, 2 some new ideas, 1 not novel |
| Technical correctness | Soundness: correct or minor errors, major errors, not applicable. Near-binary | Correctness: 4 checked in detail, 3 sound but unchecked, 2 minor fixable flaws, 1 major flaws |
| Evidence for the claims | No separate field | Claims supported by evidence, 4 very convincing down to 1 unconvincing |
| Significance | Significant, somewhat significant, not significant, not applicable | No such subscale, and the call does not name significance |
| Reproducibility | Sufficient detail to reproduce the main results, some, insufficient, or no experiments | 4 resources available and details comprehensive, down to 1 incomplete. Caps at 2 while key resources are unavailable |
| Relation to prior work | All discussed clearly, cited but less clearly, minor works missing, important works missing | No separate field |
| Writing | Clarity, plus Non-conventional Contributions | Clarity of writing |
Fields from the same two forms. The tinted row is the asymmetry to plan around: AISTATS scores whether your contribution matters, UAI whether your claims hold, and the UAI call names four criteria, omitting the reproducibility its own form scores.
Fit
Both publish what they are, in a list nobody reads
Not in the blurb about premier international conferences. In the UAI taxonomy, and in the journals AISTATS admits without review.
| UAI subject area | What it names, in its own words |
|---|---|
| Representation | Constraints, Dempster-Shafer, Description Logics, Imprecise Probabilities, Influence Diagrams |
| Models | Graphical Models, Probabilistic Circuits |
| Principles | Causality |
| Algorithms | Approximate Inference, Belief Propagation, Exact Inference, Monte Carlo Methods, Probabilistic Programming, Variational Methods |
| Learning | Causal Learning, Structure Learning |
Five of the seven top categories in the UAI 2026 subject areas, read August 2026; Applications and Theory are the other two. If your method is named here, the room that knows the name meets in August.
| Journal AISTATS admits without further review | Field |
|---|---|
| Annals of Statistics | Statistics |
| Biometrika | Statistics |
| Journal of the American Statistical Association | Statistics |
| Journal of the Royal Statistical Society Series B | Statistics |
| Journal of Machine Learning Research | Machine learning |
The AISTATS 2026 Journal-to-Conference call, read August 2026: published or accepted after 1 January 2024, nothing already at an archival conference, submitted 31 January, decided 15 February, no extra peer review. A venue tells you what it is by what it hosts on sight, and exactly one of these five is a machine learning journal. UAI has no comparable track.
The rules
Same page limit, different ways to lose without a review
| AISTATS 2026 | UAI 2026 | |
|---|---|---|
| Main text | 8 pages. References, checklist and appendices sit outside them. 9 at camera ready | 8 pages, appendices unlimited. Overlength or misformatted is desk-rejected outright |
| Author list | Final at the abstract deadline, 25 September, a week before the paper. Later changes impossible | No freeze stated |
| Anonymity | Double blind on OpenReview | Double blind. Third person about your own work, no names in acknowledgements, no identifying links |
| Preprints | arXiv allowed. No promotion while under submission, and never label it an AISTATS submission | arXiv, technical reports and non-archival workshops are not dual submissions. Promotion unmentioned |
| Reviewing obligation | One author nominated as a reviewer, up to 3 papers | One author reviews on request, if needed |
| Reviews per paper | Three to four: the call says at least 3, the guidelines say four | Not stated. Reviewers take 6 to 7 papers, chairs about 10 |
| LLMs, for authors | Permitted responsibly. Hidden prompts aimed at LLM reviewers mean rejection and a report to the board | No LLM authorship. Writing tool, yes |
| LLMs, for reviewers | Strictly prohibited, including pasting a submission into a tool | Banned for reviews and meta-reviews, except grammar checking |
| Reproducibility | A checklist, optional at submission, required at camera ready, outside the 8 pages | No checklist. One of the five scored subscales |
| Promised code | Promise a release, fail to deliver, and the FAQ says automatic rejection | Not stated |
From the AISTATS 2026 call for papers, author FAQ and area chair guidelines and the UAI 2026 call, read August 2026 on the AISTATS and UAI sites. The tinted row catches people: the AISTATS abstract deadline seals authorship, and a title or abstract that changes much afterwards is itself a desk-rejection trigger.
The number nobody has
Neither venue publishes an acceptance rate
Not on the calls, the dates pages, the FAQs, the reviewer or area chair guidelines, either homepage or either accepted-papers listing. So none appears here.
Every figure you have seen for these two is somebody's assertion
Aggregators print rates for these two because a blank cell looks broken, and the figure gets cited onward by readers who assume the venue published it. The count we stand behind is a numerator: 328 unique OpenReview forum links on the UAI 2026 accepted-papers listing, 329 PMLR paper pages, volume PMLR v337. Without a submission count that is not half of a rate. We did not count AISTATS, so no AISTATS number appears here.
Counted from the UAI 2026 accepted papers listing on 14 August 2026; the page states no total and no rate. Need one for a fellowship form? Ask the programme chairs.
Free tool, no account
Where should this paper go?
Paste the title and the abstract. It ranks the venues that publish this kind of work, with the reach and the article charge OpenAlex has on record.
The order matters more than the pick
AISTATS first, UAI as the room your reviews walk into five weeks later. Whether those are the right two rooms for this paper is the question underneath the order.
Matching needs no account and predicts no decision: what it cannot do is written down. What ends a submission without a review is on the pre-submission checklist.
phdflow reads a finished draft against the guidelines of the venue you are aiming at and returns an accept or reject call with the objections behind it. Scored against 697 real decisions: AUC 0.91 and 96% accuracy on 297 ICLR 2025 papers, one false accept in 149 rejects. The full measurement. Read a finished review before you pay for one. Your first paper gets a free preview; the full report is 60 credits, €9. No account needed. Pricing.