AISTATS Related Work

Use this to audit novelty and eligibility. Reopen the current CFP for dual-submission, anonymity, and prior-publication rules before advising authors.

Positioning checks

• Separate statistical novelty from engineering improvement: new estimator, bound, inference procedure, optimization analysis, uncertainty method, or empirical insight.
• Compare to both ML conference work and statistics literature; AISTATS reviewers often expect both communities to be represented.
• Treat PMLR, journal, and formal conference proceedings as archival unless current rules say otherwise.
• Cite arXiv and workshop versions in a way that preserves double-blind review. Do not point reviewers to identity-revealing pages.
• Explain overlap with any concurrent or prior version, and do not submit duplicate archival work.
• Use related work to sharpen what is new: assumption weakening, finite-sample behavior, computational efficiency, uncertainty calibration, robustness, or empirical regime.

Two-community coverage table

Literature lane	Typical sources	What AISTATS reviewers check
ML conferences	NeurIPS, ICML, ICLR, UAI, COLT, prior AISTATS volumes in PMLR	Whether the nearest ML method is compared or explicitly distinguished
Statistics journals	Annals of Statistics, JMLR, JASA, Biometrika, EJS	Whether classical estimators and known rates are acknowledged
Applied statistical fields	Econometrics, biostatistics, epidemiology	Whether identification and inference assumptions follow standard usage

A bibliography citing only ML venues tells a statistician reviewer that known statistical results may be getting rediscovered — a recognizable AISTATS reject pattern that no amount of benchmark strength repairs.

Positioning vignette

Imagine the paper proposes a variance-reduced off-policy evaluation estimator with an asymptotic normality result. Its nearest neighbors: a NeurIPS estimator with no inference guarantee, a JASA semiparametric efficiency bound, and a prior AISTATS paper with a slower rate. The novelty sentence should name all three contrasts — inference where the ML line had none, computational tractability where the statistics line stayed abstract, and a sharper rate than the direct predecessor.

Concurrent-work judgment calls

• Independently concurrent arXiv work: cite neutrally, state the technical difference, and avoid priority claims that reviewers cannot verify.
• Your own workshop version: typically non-archival and citable, but verify against the current CFP wording and keep the citation phrased so double-blind review survives.
• When in doubt about archival status of a venue, declare the overlap in the submission form rather than gambling on a chair's interpretation.

Output format

[Eligibility] clear / needs declaration / risky
[Closest literatures] <ML/statistics/application>
[Nearest 3 works] <work -> distinction>
[Archival-overlap risk] <none/issues>
[Novelty sentence] <AISTATS-ready contribution contrast>