BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.4.0.1//EN TZID:Europe/Paris X-WR-TIMEZONE:Europe/Paris BEGIN:VEVENT UID:190@isdm.umontpellier.fr DTSTART;TZID=Europe/Paris:20260113T140000 DTEND;TZID=Europe/Paris:20260113T140000 DTSTAMP:20260825T130844Z URL:https://isdm.umontpellier.fr/events/tabicl-a-tabular-foundation-model- for-in-context-learning-on-large-data/ SUMMARY:TabICL: A Tabular Foundation Model for In-Context Learning on Large Data DESCRIPTION:Online\n\nMachine Learning in Montpellier\, Theory &\; Pract ice - Marine Le Morvan (Inria Saclay)\n\nThe long-standing dominance of gr adient-boosted decision trees on tabular data is currently challenged by t abular foundation models using In-Context Learning (ICL): setting the trai ning data as context for the test data and predicting in a single forward pass without parameter updates. While TabPFNv2 foundation model excels on tables with up to 10K samples\, its alternating column- and row-wise atten tions make handling large training sets computationally prohibitive. So\, can ICL be effectively scaled and deliver a benefit for larger tables? We introduce TabICL\, a tabular foundation model for classification\, pretrai ned on synthetic datasets with up to 60K samples and capable of handling 5 00K samples on affordable resources. This is enabled by a novel two-stage architecture: a column-then-row attention mechanism to build fixed-dimensi onal embeddings of rows\, followed by a transformer for efficient ICL. Acr oss 200 classification datasets from the TALENT benchmark\, TabICL is on p ar with TabPFNv2 while being systematically faster (up to 10 times)\, and significantly outperforms all other approaches. On 53 datasets with over 1 0K samples\, TabICL surpasses both TabPFNv2 and CatBoost\, demonstrating t he potential of ICL for large data.\n\nMachine Learning in Montpellier\, T heory &\; Practice ATTACH;FMTTYPE=image/jpeg:https://isdm.umontpellier.fr/wp-content/uploads/ 2026/06/ml-mtp-gC78d5.png CATEGORIES:ML MTP END:VEVENT BEGIN:VTIMEZONE TZID:Europe/Paris X-LIC-LOCATION:Europe/Paris BEGIN:STANDARD DTSTART:20251026T020000 TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET END:STANDARD END:VTIMEZONE END:VCALENDAR