Publications

MultiTab: A Scalable Foundation for Multitask Learning on Tabular Data
Dimitrios Sinodinos
Jack Yi Wei
Narges Armanfard
Tabular data is the most abundant data type in the world, powering systems in finance, healthcare, e-commerce, and beyond. As tabular datase… (voir plus)ts grow and span multiple related targets, there is an increasing need to exploit shared task information for improved multitask generalization. Multitask learning (MTL) has emerged as a powerful way to improve generalization and efficiency, yet most existing work focuses narrowly on large-scale recommendation systems, leaving its potential in broader tabular domains largely underexplored. Also, existing MTL approaches for tabular data predominantly rely on multi-layer perceptron-based backbones, which struggle to capture complex feature interactions and often fail to scale when data is abundant, a limitation that transformer architectures have overcome in other domains. Motivated by this, we introduce MultiTab-Net, the first multitask transformer architecture specifically designed for large tabular data. MultiTab-Net employs a novel multitask masked-attention mechanism that dynamically models feature-feature dependencies while mitigating task competition. Through extensive experiments, we show that MultiTab-Net consistently achieves higher multitask gain than existing MTL architectures and single-task transformers across diverse domains including large-scale recommendation data, census-like socioeconomic data, and physics datasets, spanning a wide range of task counts, task types, and feature modalities. In addition, we contribute MultiTab-Bench, a generalized multitask synthetic dataset generator that enables systematic evaluation of multitask dynamics by tuning task count, task correlations, and relative task complexity. Our code is publicly available at https://github.com/Armanfard-Lab/MultiTab.
2025-11-12
ArXiv (prépublication)
doi.org
arxiv.org
The Geometry and Topology of Modular Addition Representations
Gabriela Moisescu-Pareja
Gavin McCracken
Harley Wiltzer
Colin Daniels
Vincent Létourneau
Doina Precup
Jonathan Love
The Clock and Pizza interpretations, associated with neural architectures differing in either uniform or learnable attention, were introduce… (voir plus)d to argue that different architectural designs can yield distinct circuits for modular addition. Applying geometric and topological analyses to learned representations, we show that this is not the case: Clock and Pizza circuits are topologically and geometrically equivalent and are thus equivalent representations.
2025-11-12
TAG-DS/2025/Conference (poster)
openreview.net
openreview.net
A HOT Dataset: 150,000 Buildings for HVAC Operations Transfer Research
Anaïs Berkes
Yoshua Bengio
David Rolnick
Donna Vakalis
2025-11-10
Proceedings of the 12th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation (publié)
doi.org
A HOT Dataset: 150,000 Buildings for HVAC Operations Transfer Research
Anaïs Berkes
Yoshua Bengio
David Rolnick
Donna Vakalis
About 12% of global energy consumption is attributable to heating, ventilation, and air conditioning (HVAC) systems in buildings [11]. Machi… (voir plus)ne learning-based intelligent HVAC control offers significant energy efficiency potential, but progress is constrained by limited data for training and evaluating performance across different kinds of buildings. Existing datasets primarily target energy prediction rather than control applications, forcing studies to rely on limited building sets or single-variable perturbations that fail to capture real-world complexity. We present HOT (HVAC Operations Transfer), the first large-scale open-source dataset purpose-built for research into transfer learning in building control. HOT contains 159,744 unique building-weather combinations with systematic variations across envelope properties, occupancy patterns, and climate conditions spanning all 19 ASHRAE climate zones across 76 global locations. We formalise a comprehensive similarity-based framework with quantitative metrics for assessing transfer feasibility between source and target buildings across multiple context dimensions. Our key contributions: (1) a large-scale, open dataset and tooling enabling systematic, multi-variable transfer studies across 19 climate zones; (2) a quantitative similarity framework spanning geometry, thermal, climate, and function; and (3) zero-shot climate transfer experiments showing why realistic context variation matters for HVAC control.
2025-11-10
ACM Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation (publié)
doi.org
Iterative Monte Carlo Tree Search for Neural Architecture Search
Mehraveh Javan
Matthew Toews
Marco Pedersoli
2025-11-10
Proceedings of the Fourth International Conference on Automated Machine Learning (publié)
proceedings.mlr.press
proceedings.mlr.press
LT-Soups: Bridging Head and Tail Classes via Subsampled Model Soups
Masih Aminbeidokhti
Subhankar Roy
Éric Granger
Elisa Ricci
Marco Pedersoli
Real-world datasets typically exhibit long-tailed (LT) distributions, where a few head classes dominate and many tail classes are severely u… (voir plus)nderrepresented. While recent work shows that parameter-efficient fine-tuning (PEFT) methods like LoRA and AdaptFormer preserve tail-class performance on foundation models such as CLIP, we find that they do so at the cost of head-class accuracy. We identify the head-tail ratio, the proportion of head to tail classes, as a crucial but overlooked factor influencing this trade-off. Through controlled experiments on CIFAR100 with varying imbalance ratio (
2025-11-10
ArXiv (prépublication)
doi.org
arxiv.org
Observations of AGN-driven feedback: dynamics and ionization of the filaments in M87
Camille Poitras
Marie-Lou Gendron-Marsolais
Valeria Olivares
Yuan Li
Adrien Picquenot
Aurora Simionescu
Matteo Fossati
Alessandro Boselli
Laura Hermosa Muñoz
Sara Cazzoli
Julie Hlavacek-Larrondo
Annabelle Richard-Laferrière
We present a comprehensive kinematic and ionization analysis of the warm ionized filaments (…
2025-11-10
Monthly Notices of the Royal Astronomical Society (publié)
doi.org
arxiv.org
Searching for Free-floating Planets with TESS: Results from Sectors 61–65
Michelle Kunimoto
William DeRocco
Nolan Smyth
Steve Bryson
B. Scott Gaudi
Abstract Though free-floating planets (FFPs) may outpopulate their bound counterparts in the terrestrial-mass range, they remain one of the … (voir plus)least explored exoplanet demographics. Due to their negligible electromagnetic emission at all wavelengths, the only observational technique able to detect these worlds is gravitational microlensing. Microlensing by terrestrial-mass FFPs induces rare, short-duration magnifications of background stars, requiring high-cadence, wide-field surveys to detect these events. The Transiting Exoplanet Survey Satellite (TESS), though designed to detect close-bound exoplanets via transits, boasts a full-frame image cadence as short as 200 s and has monitored hundreds of millions of stars, providing a unique dataset in which to search for rare short-duration transients. We have performed a preliminary search for FFP microlensing in 7.5 million light curves from TESS Sectors 61–65. We find one short-duration event with a light-curve morphology consistent with expectations for a low-mass FFP, but in tension with the expected FFP abundance in this mass range. We consider possible false-positive interpretations of this event such as stellar flares, heartbeat binaries, and centrifugal breakout. We find that all interpretations pose some challenges, and we discuss the possibility that the event may constitute a first example of a new class of pernicious false positives that future space-based microlensing efforts will encounter. Our ongoing search through the TESS dataset will significantly support the upcoming hunt for rogue worlds with dedicated space-based microlensing surveys, and our results may be used alongside these surveys to place interesting constraints on the spatial distribution of FFPs in the Galaxy.
2025-11-10
Astronomical Journal (publié)
doi.org
Adaptive Group Robust Ensemble Knowledge Distillation
Patrik Joslin Kenfack
Ulrich Aivodji
Samira Ebrahimi Kahou
Neural networks can learn spurious correlations in the data, often leading to performance disparity for underrepresented subgroups. Studies … (voir plus)have demonstrated that the disparity is amplified when knowledge is distilled from a complex teacher model to a relatively"simple"student model. Prior work has shown that ensemble deep learning methods can improve the performance of the worst-case subgroups; however, it is unclear if this advantage carries over when distilling knowledge from an ensemble of teachers, especially when the teacher models are debiased. This study demonstrates that traditional ensemble knowledge distillation can significantly drop the performance of the worst-case subgroups in the distilled student model even when the teacher models are debiased. To overcome this, we propose Adaptive Group Robust Ensemble Knowledge Distillation (AGRE-KD), a simple ensembling strategy to ensure that the student model receives knowledge beneficial for unknown underrepresented subgroups. Leveraging an additional biased model, our method selectively chooses teachers whose knowledge would better improve the worst-performing subgroups by upweighting the teachers with gradient directions deviating from the biased model. Our experiments on several datasets demonstrate the superiority of the proposed ensemble distillation technique and show that it can even outperform classic model ensembles based on majority voting.
2025-11-09
Transactions on Machine Learning Research (accepté)
doi.org
openreview.net
Comparability of Canadian SARS-CoV-2 seroprevalence estimates with statistical adjustment for socio-demographic representation
Yuan Yu
Jiacheng Chen
Matthew J. Knight
Sheila F. O’Brien
David L. Buckeridge
Carmen L. Charlton
W. Alton Russell
OBJECTIVE SARS-CoV-2 serological surveillance used blood donors, research cohorts, and residual patient samples. Differences in socio-demogr… (voir plus)aphic characteristics across these sources may bias seroprevalence estimates, necessitating statistical adjustment. METHODS We re-analyzed data from six serosurveillance sources, comparing the estimated percent of the population positive for SARS-CoV-2 anti-nucleocapsid antibodies for six regions during periods when the sources' sample collection overlapped. We assessed the concordance between sources with and without using multilevel regression and poststratification (MRP) to adjust for differences in representation by age, sex, and race. RESULTS Across regions and timepoints, unadjusted seroprevalence differed between sources by up to 20%. MRP did not consistently improve comparability of seroprevalence across sources. In 2022, seroprevalence was consistently highest among blood donors, and MRP increased regional seroprevalence across all sources (except in Manitoba during January-April 2022 in ABC Study). In a secondary regression analysis, immunoassay kit and sample type (dried blood spot or venous blood draw) strongly influenced the odds that a sample was classified as seropositive. CONCLUSION Adjusting for representativeness using common socio-demographic variables did not systematically improve concordance in seropositivity estimates between serosurveillance sources. While discrepancies between sources might be influenced by studies' representativeness of characteristics we did not assess, methods for measuring seropositivity appear to explain much of the differences between sources. Serosurveillance findings are influenced by many aspects of study design beyond representativeness, such as sample type (venous blood draw or dried blood spots), choice of immunoassay, and laboratory procedures such as dilution or immunoassay calibration.
2025-11-09
Canadian Journal of Public Health (publié)
doi.org
Efficient, Non‐Destructive Transfer of Wafer‐Scale Monolayer MoS
<sub>2</sub>
by Interface Engineering
Zheng Wei
Yongqing Cai
Jieying Liu
Liyan Zhang
Jiaojiao Zhao
Jian Tang
Li Li
Qinqin Wang
Huimin Zhang
Zhihua Zhang
Dongxia Shi
Luojun Du
2025-11-09
Advanced Functional Materials (publié)
doi.org
Identifying indicators of consciousness in AI systems
Patrick Butlin
Robert Long
Tim Bayne
Yoshua Bengio
Jonathan Birch
David Chalmers
Axel Constant
George Deane
Eric Elmoznino
Stephen M. Fleming
Xu Ji
Ryota Kanai
Colin Klein
Grace Lindsay
Matthias Michel
Liad Mudrik
Megan A.K. Peters
Eric Schwitzgebel
Jonathan Simon
Rufin Vanrullen
2025-11-09
Trends in Cognitive Sciences (publié)
doi.org