Eight universities share the lead with two top-paper wins since 2024

Across the verified Best Paper and Outstanding Paper winners at AAAI, IJCAI, ICLR, ICML, NeurIPS, ACL and CVPR, no university appears on more than two distinct winning papers. The count credits every university represented in a paper’s author affiliations, so one paper may count for several universities.

2
leading wins per university
8
co-leading universities
16
university–paper links
7
conference series scanned

Top-paper wins credited by author affiliation

Distinct papers · 2024–24 Aug 2026
Carnegie Mellon
2
Harvard
2
National Univ. of Singapore
2
Princeton
2
Stanford
2
Tel Aviv Univ.
2
Univ. of Oxford
2
Univ. of Toronto
2

What drives the tie

01
ICML supplies half of the links. Eight of the 16 university–paper credits come from ICML winners in 2024 or 2025, including joint university–industry papers.
02
Shared papers create shared leaders. NeurIPS 2025’s Artificial Hivemind counts for both Carnegie Mellon and Stanford; ICML 2024’s Information Complexity paper counts for both Tel Aviv and Toronto.
03
Oxford’s two wins are both vision papers. VGGT won at CVPR 2025 and D4RT at CVPR 2026; the latter’s short and full titles were deduplicated as one paper.

The winning papers

UniversityPaperVenueAwardPrimary author affiliations
Carnegie MellonRoll the Dice & Look Before You LeapICML 2025Outstanding PaperGoogle Research; Carnegie Mellon University
Carnegie MellonArtificial HivemindNeurIPS 2025Best PaperUniversity of Washington; Carnegie Mellon; Ai2; Lila Sciences; Stanford
HarvardThe Value of Prediction in Identifying the Worst-OffICML 2025Outstanding PaperLMU Munich; Munich Center for Machine Learning; Harvard University
HarvardTrain for the Worst, Plan for the BestICML 2025Outstanding PaperUniversity of Texas at Austin; Harvard University
National Univ. of SingaporeStochastic Taylor Derivative EstimatorNeurIPS 2024Best PaperNUS Computing; Sea AI Lab
National Univ. of SingaporeAlphaEditICLR 2025Outstanding PaperUniversity of Science and Technology of China; National University of Singapore
PrincetonSafety Alignment Should Be Made More Than Just a Few Tokens DeepICLR 2025Outstanding PaperPrinceton University; Google DeepMind
PrincetonConformal Prediction as Bayesian QuadratureICML 2025Outstanding PaperPrinceton University
StanfordCollabLLMICML 2025Outstanding PaperStanford University; Microsoft; Georgia Tech
StanfordArtificial HivemindNeurIPS 2025Best PaperUniversity of Washington; Carnegie Mellon; Ai2; Lila Sciences; Stanford
Tel Aviv UniversityNever Train from ScratchICLR 2024Outstanding PaperTel Aviv University; IBM Research
Tel Aviv UniversityInformation Complexity of Stochastic Convex OptimizationICML 2024Best PaperBen-Gurion; Vector Institute; Google DeepMind; Northeastern; Tel Aviv; Toronto
University of OxfordVGGT: Visual Geometry Grounded TransformerCVPR 2025Best PaperUniversity of Oxford; Meta AI
University of OxfordEfficiently Reconstructing Dynamic Scenes One D4RT at a TimeCVPR 2026Best PaperGoogle DeepMind; University College London; University of Oxford
University of TorontoInformation Complexity of Stochastic Convex OptimizationICML 2024Best PaperBen-Gurion; Vector Institute; Google DeepMind; Northeastern; Tel Aviv; Toronto
University of TorontoEvery Bit Helps: Achieving the Optimal Distortion with a Few QueriesAAAI 2025Outstanding PaperUniversity of Toronto
Method: web search and paper-page extraction run 24 Aug 2026 across seven conference series. Counts use distinct winning papers and credit each university listed among the authors’ primary affiliations; companies and standalone institutes are not ranked. Honorable mentions, student, demo, dataset, benchmark and test-of-time awards were excluded. 2026 is year-to-date.
Keenable · made with SELECT* · 4,245 pages in 4m 27s · Open the chatShare: XLinkedInReddit