Publications

Whither symbols in the era of advanced neural networks?
Thomas L. Griffiths
Brenden M. Lake
R. Thomas McCoy
Ellie Pavlick
Taylor Webb
Some of the strongest evidence that human minds should be thought about in terms of symbolic systems has been the way they combine ideas, pr… (see more)oduce novelty, and learn quickly. We argue that modern neural networks -- and the artificial intelligence systems built upon them -- exhibit similar abilities. This undermines the argument that the cognitive processes and representations used by human minds are symbolic, although the fact that these neural networks are typically trained on data generated by symbolic systems illustrates that such systems play an important role in characterizing the abstract problems that human minds have to solve. This argument leads us to offer a new agenda for research on the symbolic basis of human thought.
2025-08-06
ArXiv (preprint)
doi.org
arxiv.org
Persistent Instability in LLM's Personality Measurements: Effects of Scale, Reasoning, and Conversation History
Tommaso Tosato
Saskia Helbling
Yorguin-Jose Mantilla-Ramos
Mahmood Hegazy
Alberto Tosato
D. Lemay
Irina Rish
Guillaume Dumas
2025-08-05
ArXiv (preprint)
doi.org
arxiv.org
Manifold Filter-Combine Networks
Joyce Chew
Edward De Brouwer
Smita Krishnaswamy
Deanna Needell
Michael Perlmutter
In order to better understand manifold neural networks (MNNs), we introduce Manifold Filter-Combine Networks (MFCNs). Our filter-combine fra… (see more)mework parallels the popular aggregate-combine paradigm for graph neural networks (GNNs) and naturally suggests many interesting families of MNNs which can be interpreted as manifold analogues of various popular GNNs. We propose a method for implementing MFCNs on high-dimensional point clouds that relies on approximating an underlying manifold by a sparse graph. We then prove that our method is consistent in the sense that it converges to a continuum limit as the number of data points tends to infinity, and we numerically demonstrate its effectiveness on real-world and synthetic data sets.
2025-08-04
Sampling Theory, Signal Processing, and Data Analysis (published)
doi.org
arxiv.org
Single-nucleus chromatin accessibility profiling identifies cell types and functional variants contributing to major depression
Anjali Chawla
Doruk Cakmakci
Laura M. Fiori
Wenmin Zang
Malosree Maitra
Jennie Yang
Dariusz Żurawek
Gabriella Frosi
Reza Rahimian
Haruka Mitsuhashi
Maria Antonietta Davoli
MA Davoli
Ryan Denniston
Gary Gang Chen
Volodymyr Yerko
Deborah Mash
Kiran Girdhar
Schahram Akbarian
Naguib Mechawar
Matthew Suderman … (see 3 more)
Yuemei Li
Corina Nagy
Gustavo Turecki
2025-08-04
Nature Genetics (published)
doi.org
Understanding In-Context Learning of Linear Models in Transformers Through an Adversarial Lens
Usman Anwar
Johannes Von Oswald
Louis Kirsch
David M. Krueger
Spencer Frei
In this work, we make two contributions towards understanding of in-context learning of linear models by transformers. First, we investigate… (see more) the adversarial robustness of in-context learning in transformers to hijacking attacks — a type of adversarial attacks in which the adversary’s goal is to manipulate the prompt to force the transformer to generate a specific output. We show that both linear transformers and transformers with GPT-2 architectures are vulnerable to such hijacking attacks. However, adversarial robustness to such attacks can be significantly improved through adversarial training --- done either at the pretraining or finetuning stage --- and can generalize to stronger attack models. Our second main contribution is a comparative analysis of adversarial vulnerabilities across transformer models and other algorithms for learning linear models. This reveals two novel findings. First, adversarial attacks transfer poorly between larger transformer models trained from different seeds despite achieving similar in-distribution performance. This suggests that transformers of the same architecture trained according to the same recipe may implement different in-context learning algorithms for the same task. Second, we observe that attacks do not transfer well between classical learning algorithms for linear models (single-step gradient descent and ordinary least squares) and transformers. This suggests that there could be qualitative differences between the in-context learning algorithms that transformers implement and these traditional algorithms.
2025-08-04
TMLR (accepted)
openreview.net
openreview.net
CellForge: Agentic Design of Virtual Cell Models
Xiangru Tang
Zhuoyun Yu
Jiapeng Chen
Yan Cui
Yanjun Shao
Weixu Wang
Fang Wu
Yuchen Zhuang
Wenqi Shi
Zhi Huang
Arman Cohan
Xihong Lin
Fabian Theis
Smita Krishnaswamy
Mark B. Gerstein
Virtual cell modeling represents an emerging frontier at the intersection of artificial intelligence and biology, aiming to predict quantiti… (see more)es such as responses to diverse perturbations quantitatively. However, autonomously building computational models for virtual cells is challenging due to the complexity of biological systems, the heterogeneity of data modalities, and the need for domain-specific expertise across multiple disciplines. Here, we introduce CellForge, an agentic system that leverages a multi-agent framework that transforms presented biological datasets and research objectives directly into optimized computational models for virtual cells. More specifically, given only raw single-cell multi-omics data and task descriptions as input, CellForge outputs both an optimized model architecture and executable code for training virtual cell models and inference. The framework integrates three core modules: Task Analysis for presented dataset characterization and relevant literature retrieval, Method Design, where specialized agents collaboratively develop optimized modeling strategies, and Experiment Execution for automated generation of code. The agents in the Design module are separated into experts with differing perspectives and a central moderator, and have to collaboratively exchange solutions until they achieve a reasonable consensus. We demonstrate CellForge's capabilities in single-cell perturbation prediction, using six diverse datasets that encompass gene knockouts, drug treatments, and cytokine stimulations across multiple modalities. CellForge consistently outperforms task-specific state-of-the-art methods. Overall, CellForge demonstrates how iterative interaction between LLM agents with differing perspectives provides better solutions than directly addressing a modeling challenge. Our code is publicly available at https://github.com/gersteinlab/CellForge.
2025-08-03
ArXiv (preprint)
doi.org
arxiv.org
Infrared Object Detection with Ultra Small ConvNets: Is ImageNet Pretraining Still Useful?
Srikanth Muralidharan
Heitor Rapela Medeiros
Masih Aminbeidokhti
Éric Granger
Marco Pedersoli
Many real-world applications require recognition models that are robust to different operational conditions and modalities, but at the same … (see more)time run on small embedded devices, with limited hardware. While for normal size models, pre-training is known to be very beneficial in accuracy and robustness, for small models, that can be employed for embedded and edge devices, its effect is not clear. In this work, we investigate the effect of ImageNet pretraining on increasingly small backbone architectures (ultra-small models, with
2025-08-03
ArXiv (preprint)
doi.org
arxiv.org
The impact of gender on pediatric surgical access and outcomes in Africa
Sacha Williams
Olivia Serhan
Jenny Wang
Christian Guindi,
Elena Guadagno
Maeve Trudeau
Emmanuel Ameh
Kokila Lakhoo
Dan Poenaru
Girls, whose care is often affected by barriers steeped in gender inequity, may be at higher risk of poor surgical outcomes. This study expl… (see more)ored the impact of gender on pediatric surgical care in Africa. Differences in access to care and clinical outcomes for boys and girls were examined for pediatric surgical conditions that do not differ by physiological sex. A systematic review of African pediatric surgical studies ensued, followed by a random effects meta-analysis, and risk of bias assessment. Of the 12281 records retrieved, 54 were selected for review. Most studies were retrospective (57.4%), single-site (94.4%), from Egypt, Nigeria, Ghana, or Ethiopia (55.6%), focussed on gastrointestinal conditions (63.0%), published in 2010 or sooner (85.1%), had study durations of 5 years or less (68.5%), and cohorts of less than 200 children (57.4%). Sixty percent reported the outcome of mortality. Meta-analysis odds ratios revealed surgery was performed 3.6 times more often on boys (95% CI: 2.6, 4.9); and mortality was 1.6 times greater for girls (95% CI: 1.3, 2.0). African girls appear to face gender inequities in pediatric surgical care. Findings will be further explored in a mixed-methods study. I Gender disparities in global surgical care have been documented in the African adult population. However gender specific differentials in surgical access and outcomes have yet to be documented for African pediatric populations. This study provides first-time evidence of gender inequity in pediatric surgical care in Africa.
2025-08-03
Pediatric Surgery International (published)
doi.org
A Guide to Misinformation Detection Data and Evaluation
Camille Thibault
Jacob-Junqi Tian
Gabrielle Peloquin-Skulski
Taylor Lynn Curtis
James Zhou
Florence Laflamme
Luke Yuxiang Guan
Reihaneh Rabbany
Jean-François Godbout
Kellin Pelrine
Misinformation is a complex societal issue, and mitigating solutions are difficult to create due to data deficiencies. To address this, we h… (see more)ave curated the largest collection of (mis)information datasets in the literature, totaling 75. From these, we evaluated the quality of 36 datasets that consist of statements or claims, as well as the 9 datasets that consist of data in purely paragraph form. We assess these datasets to identify those with solid foundations for empirical work and those with flaws that could result in misleading and non-generalizable results, such as spurious correlations, or examples that are ambiguous or otherwise impossible to assess for veracity. We find the latter issue is particularly severe and affects most datasets in the literature. We further provide state-of-the-art baselines on all these datasets, but show that regardless of label quality, categorical labels may no longer give an accurate evaluation of detection model performance. Finally, we propose and highlight Evaluation Quality Assurance (EQA) as a tool to guide the field toward systemic solutions rather than inadvertently propagating issues in evaluation. Overall, this guide aims to provide a roadmap for higher quality data and better grounded evaluations, ultimately improving research in misinformation detection. All datasets and other artifacts are available at https://misinfo-datasets.complexdatalab.com/.
2025-08-02
ACM SIGKDD Conference on Knowledge Discovery and Data Mining (published)
doi.org
openreview.net
Responsible AI Day
Ebrahim Bagheri
Faezeh Ensan
Calvin Hillis
Reihaneh Rabbany
Robin Cohen
Benjamin C. M. Fung
Sébastien Gambs
This special day event on Responsible Artificial Intelligence (AI) brings together researchers, practitioners, and policymakers to explore h… (see more)ow data mining and machine learning systems can be designed to align with ethical principles, societal values, and human well-being. As AI technologies increasingly influence decisions in healthcare, finance, governance, and social systems, there is a critical need to develop frameworks that embed fairness, accountability, and privacy directly into the foundations of knowledge discovery. This full-day event will feature a mix of invited talks, interactive debates, expert panels, and peer-reviewed research presentations, all focused on the practical integration of ethical design into data-driven systems. The Responsible AI Day builds on the success of Canada's NSERC CREATE Program on Responsible AI, an interdisciplinary initiative training the next generation of AI researchers across computer science, law, bioethics, public health, and media studies. Topics will span scalable AI governance, privacy-preserving computation, algorithmic bias mitigation, and the socio-legal tensions emerging in generative AI. By positioning responsible AI as a sociotechnical challenge, this special day aligns with KDD's mission of advancing data science that is not only technically robust but also socially conscious.
2025-08-02
Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (published)
doi.org
Temporal Graph Learning Workshop
Shenyang Huang
Daniele Zambon
Andrea Cini
Farimah Poursafaei
Jacob Chmura
Julia Gastinger
Reihaneh Rabbany
Michael M. Bronstein
2025-08-02
ACM SIGKDD Conference on Knowledge Discovery and Data Mining (published)
doi.org
Unified Game Moderation: Soft-Prompting and LLM-Assisted Label Transfer for Resource-Efficient Toxicity Detection
Zachary Yang
Domenico Tullo
Reihaneh Rabbany
Toxicity detection in gaming communities faces significant scaling challenges when expanding across multiple games and languages, particular… (see more)ly in real-time environments where computational efficiency is crucial. We present two key findings to address these challenges while building upon our previous work on ToxBuster, a BERT-based real-time toxicity detection system. First, we introduce a soft-prompting approach that enables a single model to effectively handle multiple games by incorporating game-context tokens, matching the performance of more complex methods like curriculum learning while offering superior scalability. Second, we develop an LLM-assisted label transfer framework using GPT-4o-mini to extend support to seven additional languages. Evaluations on real game chat data across French, German, Portuguese, and Russian achieve macro F1-scores ranging from 32.96% to 58.88%, with particularly strong performance in German, surpassing the English benchmark of 45.39%. In production, this unified approach significantly reduces computational resources and maintenance overhead compared to maintaining separate models for each game and language combination. At Ubisoft, this model successfully identifies an average of 50 players, per game, per day engaging in sanctionable behavior.
2025-08-02
ACM SIGKDD Conference on Knowledge Discovery and Data Mining (published)
doi.org
arxiv.org