Publications

FairFLRep: Fairness aware fault localization and repair of Deep Neural Networks
Moses Openja
Paolo Arcaini
Foutse Khomh
Fuyuki Ishikawa
2025-09-09
ACM Transactions on Software Engineering and Methodology (publié)
doi.org
arxiv.org
EZH2 Inhibition Induces an Integrated Stress Response Driving Glutamine-Dependent Vulnerability in TNBC
Lucas Porras
Marina Fukano
Ann-Sophie Gironne
Elise Quadri
Gabriel Alzial
Hugo Philippeau
Yousef Aleassa
Anie Monast
Faustine Gorse
Myriame Saint-Arnaud
Mariana De Sa Tavares Russo
Sylvie Mader
Daina Avizonis
Sébastien Lemieux
Morag Park
Geneviève Deblois
EZH2, the catalytic subunit of Polycomb Repressive Complex II, is highly expressed and associated with poor prognosis in triple-negative bre… (voir plus)ast cancer (TNBC). Despite inducing significant changes in chromatin profiles and gene expression, EZH2 inhibition in TNBC models has limited impact on growth, suggesting adaptive compensatory mechanisms. Here, we demonstrate that EZH2 inhibition induces accumulation of double-stranded RNA and misfolded proteins in TNBC, activating an integrated stress response (ISR) via the PKR/PERK-eIF2α pathway. We identify Activating Transcription Factor 4 (ATF4) as a key effector upon EZH2 inhibition, driving metabolic changes characterized by increased amino acid uptake and glutamine dependency. Targeting this ISR-ATF4-mediated metabolic response using glutaminase inhibitor in combination with EZH2 inhibition significantly impairs TNBC cell proliferation and tumor progression. These findings reveal a stress-driven metabolic adaptation that enables TNBC survival upon EZH2 blockade, highlighting inhibition of this pathway as a strategy to enhance the efficacy of EZH2 inhibitors in TNBC.
2025-09-08
bioRxiv (prépublication)
doi.org
RL Fine-Tuning Heals OOD Forgetting in SFT
Hangzhan Jin
Sitao Luan
Sicheng Lyu
Guillaume Rabusseau
Reihaneh Rabbany
Doina Precup
Mohammad Hamdaqa
The two-stage fine-tuning paradigm of Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL) has empirically shown better reas… (voir plus)oning performance than one-stage SFT for the post-training of Large Language Models (LLMs). However, the evolution and mechanism behind the synergy of SFT and RL are still under-explored and inconclusive. In our study, we find the well-known claim "SFT memorizes, RL generalizes" is over-simplified, and discover that: (1) OOD performance peaks at the early stage of SFT and then declines (OOD forgetting), the best SFT checkpoint cannot be captured by training/test loss; (2) the subsequent RL stage does not generate fundamentally better OOD capability, instead it plays an \textbf{OOD restoration} role, recovering the lost reasoning ability during SFT; (3) The recovery ability has boundaries, \ie{} \textbf{if SFT trains for too short or too long, RL cannot recover the lost OOD ability;} (4) To uncover the underlying mechanisms behind the forgetting and restoration process, we employ SVD analysis on parameter matrices, manually edit them, and observe their impacts on model performance. Unlike the common belief that the shift of model capacity mainly results from the changes of singular values, we find that they are actually quite stable throughout fine-tuning. Instead, the OOD behavior strongly correlates with the \textbf{rotation of singular vectors}. Our findings re-identify the roles of SFT and RL in the two-stage fine-tuning and discover the rotation of singular vectors as the key mechanism. %reversing the rotations induced by SFT, which shows recovery from forgetting, whereas imposing the SFT parameter directions onto a RL-tuned model results in performance degradation. Code is available at https://github.com/xiaodanguoguo/RL_Heals_SFT
2025-09-07
arXiv (Cornell University) (prépublication)
doi.org
arxiv.org
An AI system to help scientists write expert-level empirical software
Eser Aygün
Anastasiya Belyaeva
Gheorghe Comanici
Marc Coram
Hao Cui
Jake Garrison
Renee Johnston Anton Kast
Cory Y. McLean
Peter C. Norgaard
Zahra Shamsi
David Smalling
James Thompson
Subhashini Venugopalan
Brian P Williams
Chujun He
Sarah Martinson
Martyna Plomecka
Lai Wei
Yuchen Zhou
Qian-Ze Zhu … (voir 21 de plus)
Matthew Abraham
Erica Brand
Anna Bulanova
Jeffrey A. Cardille
Chris Co
Scott Ellsworth
Grace Joseph
Malcolm Kane
Ryan K. Krueger
Johan Kartiwa
D. Liebling
Jan-Matthis Lueckmann
Paul Raccuglia
Xuefei Wang
Katherine Chou
James Manyika
Yossi Matias
J.C. Platt
Lizzie Dorfman
Shibl Mourad
Michael P. Brenner
The cycle of scientific discovery is frequently bottlenecked by the slow, manual creation of software to support computational experiments. … (voir plus)To address this, we present an AI system that creates expert-level scientific software whose goal is to maximize a quality metric. The system uses a Large Language Model (LLM) and Tree Search (TS) to systematically improve the quality metric and intelligently navigate the large space of possible solutions. The system achieves expert-level results when it explores and integrates complex research ideas from external sources. The effectiveness of tree search is demonstrated across a wide range of benchmarks. In bioinformatics, it discovered 40 novel methods for single-cell data analysis that outperformed the top human-developed methods on a public leaderboard. In epidemiology, it generated 14 models that outperformed the CDC ensemble and all other individual models for forecasting COVID-19 hospitalizations. Our method also produced state-of-the-art software for geospatial analysis, neural activity prediction in zebrafish, time series forecasting and numerical solution of integrals. By devising and implementing novel solutions to diverse tasks, the system represents a significant step towards accelerating scientific progress.
2025-09-07
ArXiv (prépublication)
doi.org
arxiv.org
Discrete Audio Tokens: More Than a Survey!
Pooneh Mousavi
Gallil Maimon
Adel Moumen
Darius Petermann
Jiatong Shi
Haibin Wu
Haici Yang
Anastasia Kuznetsova
Artem Ploujnikov
Ricard Marxer
Bhuvana Ramabhadran
Benjamin Elizalde
Loren Lugosch
Jinyu Li
Yusuf Cem Sübakan
Phil Woodland
Minje Kim
Hung-yi Lee
Shinji Watanabe
Yossi Adi … (voir 1 de plus)
Mirco Ravanaelli
Discrete audio tokens are compact representations that aim to preserve perceptual quality, phonetic content, and speaker characteristics whi… (voir plus)le enabling efficient storage and inference, as well as competitive performance across diverse downstream tasks. They provide a practical alternative to continuous features, enabling the integration of speech and audio into modern large language models (LLMs). As interest in token-based audio processing grows, various tokenization methods have emerged, and several surveys have reviewed the latest progress in the field. However, existing studies often focus on specific domains or tasks and lack a unified comparison across various benchmarks. This paper presents a systematic review and benchmark of discrete audio tokenizers, covering three domains: speech, music, and general audio. We propose a taxonomy of tokenization approaches based on encoder-decoder, quantization techniques, training paradigm, streamability, and application domains. We evaluate tokenizers on multiple benchmarks for reconstruction, downstream performance, and acoustic language modeling, and analyze trade-offs through controlled ablation studies. Our findings highlight key limitations, practical considerations, and open challenges, providing insight and guidance for future research in this rapidly evolving area. For more information, including our main results and tokenizer database, please refer to our website: https://poonehmousavi.github.io/dates-website/.
2025-09-06
TMLR (accepté)
doi.org
openreview.net
Task Robustness via Re-Labelling Vision-Action Robot Data
Artur Kuramshin
Özgür Aslan
Cyrus Neary
Glen Berseth
2025-09-05
robot-learning.org/CoRL/2025/Workshop/Robot_Data (publié)
openreview.net
openreview.net
Massive Extremely High-velocity Outflow in the Quasar J164653.72+243942.2
Paola Rodríguez Hidalgo
Hyunseop Choi (최현섭)
Patrick B. Hall
Karen M. Leighly
Liliana Flores
Mikel M. Charles
Cora DeFrancesco
Julie Hlavacek-Larrondo
Laurence Perreault-Levasseur
We present the analysis of one of the most extreme quasar outflows found to date in our survey of extremely high velocity outflows (EHVO). J… (voir plus)164653.72+243942.2 (z ~ 3.04) shows variable CIV1548,1551 absorption at speeds larger than 0.1c, accompanied by SiIV, NV and Lya, and disappearing absorption at lower speeds. We perform absorption measurements using the Apparent Optical Depth method and SimBAL. We find the absorption to be very broad (Δv ~35,100 km/s in the first epoch and ~13,000 km/s in the second one) and fast (vmax ~ -50,200 km/s and -49,000 km/s, respectively). We measure large column densities (
2025-09-03
The Astrophysical Journal (publié)
doi.org
arxiv.org
MISTRAL: a model for AGN winds from radiatively efficient accretion in cosmological simulations
Marion Farcy
Michaela Hirschmann
Rachel S. Somerville
Ena Choi
Sophie Koudmani
Thorsten Naab
Rainer Weinberger
Jake S. Bennett
Aklant K. Bhowmick
Hyunseop Choi
Lars Hernquist
Julie Hlavacek-Larrondo
Bryan A. Terrazas
Francesco Valentino
ABSTRACT Feedback from active galactic nuclei (AGNs) is crucial for regulating galaxy evolution. Motivated by observations of broad absorpti… (voir plus)on line winds from rapidly accreting supermassive black holes (SMBHs), we introduce the mistral AGN feedback model, implemented in the arepo code. mistral comes in two versions: continuous radial (mistral-continuous) and stochastic bipolar momentum deposition (mistral-stochastic). Using the framework of the IllustrisTNG simulations, we explore the effect of mistral on BH and galaxy properties, through an idealized Milky Way-mass galaxy and cosmological zoom simulations run down to
2025-09-03
Monthly Notices of the Royal Astronomical Society (publié)
doi.org
arxiv.org
Behaviour Discovery and Attribution for Explainable Reinforcement Learning
Rishav
Somjit Nath
Vincent Michalski
S Ebrahimi Kahou
2025-09-01
Transactions on Machine Learning Research (accepté)
doi.org
openreview.net
Learning Laplacian Eigenvectors: a Pre-training Method for Graph Neural Networks
Howard Dai
Nyambura Njenga
Catherine Ma
Siddharth Viswanath
Ryan Pellico
Ian Adelstein
Smita Krishnaswamy
2025-09-01
ArXiv (prépublication)
arxiv.org
arxiv.org
Early Deforestation Detection in the Tropics using L-band SAR and Optical multi-sensor data and Bayesian Statistics
Africa I. Flores-Anderson
Jeffrey A. Cardille
Josef Kellndorfer
Franz J. Meyer
Pontus Olofsson
2025-08-31
International Journal of Applied Earth Observation and Geoinformation (publié)
doi.org
Metabolic Control and Frequency of Clinical Monitoring Among Canadian Children With Phenylalanine Hydroxylase Deficiency: A Retrospective Cohort Study
Nataliya Yuskiv
Ammar Saad
Beth K. Potter
Sylvia Stockler‐Ipsiroglu
John J. Mitchell
Steven Hawken
Kylie Tingley
Michael Pugliese
Monica Lamoureux
Andrea J. Chow
Jonathan B. Kronick
Kumanan Wilson
Annette Feigenbaum
Sharan Goobie
Michal Inbar-Feigenberg
Julian Little
Saadet Mercimek‐Andrews
Amy Pender
Chitra Prasad
Andreas Schulze … (voir 9 de plus)
Yannis Trakadis
Gloria Ho
Hilary Vallance
Valerie Austin
Anthony Vandersteen
Andrea C. Yu
Cheryl Rockman‐Greenberg
Aizeddin Mhanni
Pranesh Chakraborty
2025-08-31
JIMD Reports (publié)
doi.org