Use when deciding whether a project is a strong AISTATS fit, comparing AISTATS with NeurIPS, ICML, ICLR, UAI, COLT, JMLR, statistics journals, or application venues, identifying the statistical primitive of the contribution, and sharpening the AI-statistics framing before writing begins.
SkillScore58Maintained
Automatic
◆Research pack
Last verified 8/15/2026
What it changes
Passed 13 of 23 test cases with the skill, and 10 of 23 without it.
Without skill42%
With skill58%
Checks it passed that the model alone missed
The output specifies "[Best venue] UAI".
The output specifies "[Fit] strong AISTATS".
The output specifies "[Fit] better elsewhere".
and 4 more in the full report
10 fixed4 passed either way8 still failing1 broke
Pass-rate lift
+17 pp
Tokens
+19%
Agent turns
0%
Response time
-35%
Model gemini-3.6-flashMeasured 8/15/2026v114 of 23 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 10
Prompt
We conducted a multi-GPU empirical study scaling transformer models up to 70B parameters, evaluating zero-shot performance across 15 NLP benchmarks without formal statistical proofs. Evaluate conference fit and output: fit classification, target venue, one-sentence contribution, top rejection risk, and next step.
What was checked, without and with the skill
Checked
Without skill
With skill
The output specifies "[Fit] better elsewhere".
Missed
Passed
The response recommends broad ML venues such as NeurIPS, ICML, or ICLR for large-scale empirical systems work.
Passed
Passed
case-10 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md60 linessingle file
AISTATS Topic Selection
Use this before writing. AISTATS is strongest for work at the intersection of artificial
intelligence, machine learning, and statistics, especially when statistical reasoning is not
merely an evaluation detail.
Fit test
Prefer AISTATS when the contribution advances statistical foundations, inference,
uncertainty, causal or probabilistic modeling, learning theory, optimization, or empirical
methodology with clear AI/ML relevance.
Route to ICML, NeurIPS, or ICLR if the main contribution is broad ML systems, representation
learning, scaling, or deep learning practice with limited statistical novelty.
Route to UAI if the contribution is primarily uncertainty, probabilistic graphical models,
causality, decision making under uncertainty, or Bayesian reasoning.
Route to COLT if the contribution is mainly formal learning theory and the empirical story
is secondary.
Route to a statistics journal when the work needs journal-length exposition, extensive
proofs, or a statistics audience more than an AI conference audience.
Check early whether the result can be made convincing in an 8-page submission body.
Fit signal table
Signal in the project
AISTATS reading
Consistency, minimax rate, regret, or coverage result paired with experiments
Core fit — the house genre
Bayesian, causal, kernel, or high-dimensional methodology with guarantees
Core fit
Deep architecture with strong benchmarks but thin theory
Better served at NeurIPS, ICML, or ICLR
Pure theory with no plausible experiment
COLT or a statistics journal
Probabilistic reasoning without a learning angle
UAI or a statistics venue
Vignette: where a debiased estimator goes
A project delivers a debiased lasso variant with valid confidence intervals in high
dimensions and simulations confirming coverage. AISTATS reading: strong fit — an inference
guarantee plus validating experiments is exactly what this venue rewards. Strip the
inference theory and keep only prediction benchmarks, and the same project belongs at a
general ML venue; grow it into journal-length asymptotic refinements, and Annals of
Statistics or JMLR becomes the better home.