Publications

Robust Fine-Tuning from Non-Robust Pretrained Models: Mitigating Suboptimal Transfer With Adversarial Scheduling
Jonas Ngnawe
Maxime Heuillet
Sabyasachi Sahoo
Yann Batiste Pequignot
Ola Ahmad
Audrey Durand
Frederic Precioso
Christian Gagné
Fine-tuning pretrained models is a standard and effective workflow in modern machine learning. However, robust fine-tuning (RFT), which aims… (voir plus) to simultaneously achieve adaptation to a downstream task and robustness to adversarial examples, remains challenging. Despite the abundance of non-robust pretrained models in open-source repositories, their potential for RFT is less understood. We address this knowledge gap by systematically examining RFT from such non-robust models. Our experiments reveal that fine-tuning non-robust models with a robust objective, even under small perturbations, can lead to poor performance, a phenomenon that we dub \emph{suboptimal transfer}. In challenging scenarios (eg, difficult tasks, high perturbation), the resulting performance can be so low that it may be considered a transfer failure. We find that fine-tuning using a robust objective impedes task adaptation at the beginning of training and eventually prevents optimal transfer. However, we propose a novel heuristic, \emph{Epsilon-Scheduling}, a schedule over perturbation strength used during training that promotes optimal transfer. Additionally, we introduce \emph{expected robustness}, a metric that captures performance across a range of perturbations, providing a more comprehensive evaluation of the accuracy-robustness trade-off for diverse models at test time. Extensive experiments on a wide range of configurations (six pretrained models and five datasets) show that \emph{Epsilon-Scheduling} successfully prevents \emph{suboptimal transfer} and consistently improves expected robustness.
2025-09-27
ArXiv (prépublication)
arxiv.org
arxiv.org
Tracing the Representation Geometry of Language Models from Pretraining to Post-training
Melody Zixuan Li
Kumar Krishna Agrawal
Arna Ghosh
Komal K. Teru
Adam Santoro
Guillaume Lajoie
Blake A. Richards
Standard training metrics like loss fail to explain the emergence of complex capabilities in large language models. We take a spectral appro… (voir plus)ach to investigate the geometry of learned representations across pretraining and post-training, measuring effective rank (RankMe) and eigenspectrum decay (
2025-09-27
ArXiv (prépublication)
arxiv.org
arxiv.org
Active Attacks: Red-teaming LLMs via Adaptive Environments
Taeyoung YUN
Pierre-Luc St-Charles
Jinkyoo Park
Yoshua Bengio
Minsu Kim
We address the challenge of generating diverse attack prompts for large language models (LLMs) that elicit harmful behaviors (e.g., insults,… (voir plus) sexual content) and are used for safety fine-tuning. Rather than relying on manual prompt engineering, attacker LLMs can be trained with reinforcement learning (RL) to automatically generate such prompts using only a toxicity classifier as a reward. However, capturing a wide range of harmful behaviors is a significant challenge that requires explicit diversity objectives. Existing diversity-seeking RL methods often collapse to limited modes: once high-reward prompts are found, exploration of new regions is discouraged. Inspired by the active learning paradigm that encourages adaptive exploration, we introduce \textit{Active Attacks}, a novel RL-based red-teaming algorithm that adapts its attacks as the victim evolves. By periodically safety fine-tuning the victim LLM with collected attack prompts, rewards in exploited regions diminish, which forces the attacker to seek unexplored vulnerabilities. This process naturally induces an easy-to-hard exploration curriculum, where the attacker progresses beyond easy modes toward increasingly difficult ones. As a result, Active Attacks uncovers a wide range of local attack modes step by step, and their combination achieves wide coverage of the multi-mode distribution. Active Attacks, a simple plug-and-play module that seamlessly integrates into existing RL objectives, unexpectedly outperformed prior RL-based methods -- including GFlowNets, PPO, and REINFORCE -- by improving cross-attack success rates against GFlowNets, the previous state-of-the-art, from 0.07% to 31.28% (a relative gain greater than
2025-09-26
ArXiv (prépublication)
arxiv.org
arxiv.org
Continual Pre-training of MoEs: How robust is your router?
Benjamin Therien
Charles-Etienne Joseph
Zain Sarwar
Ashwinee Panda
Anirban Das
Shi-Xiong Zhang
Stephen Rawls
Sambit Sahu
Eugene Belilovsky
Irina Rish
Sparsely-activated Mixture of Experts (MoE) transformers are promising architectures for foundation models. Compared to dense transformers t… (voir plus)hat require the same amount of floating-point operations (FLOPs) per forward pass, MoEs benefit from improved sample efficiency at training time and achieve much stronger performance. Many closed-source and open-source frontier language models have thus adopted an MoE architecture. Naturally, practitioners will want to extend the capabilities of these models with large amounts of newly collected data without completely re-training them. Prior work has shown that a simple combination of replay, learning rate re-warming, and re-decaying can enable the continual pre-training (CPT) of dense decoder-only transformers with minimal performance degradation compared to full re-training. In the case of decoder-only MoE transformers, however, it is unclear how the routing algorithm will impact continual pre-training performance: 1) *do the MoE transformer's routers exacerbate forgetting relative to a dense model?*; 2) *do the routers maintain a balanced load on previous distributions after CPT?*; 3) *are the same strategies applied to dense models sufficient to continually pre-train MoE LLMs?* In what follows, we conduct a large-scale study training a 500M parameter dense transformer and four 500M-active/2B-total parameter MoE transformers, following the Switch Transformer architecture and a granular DeepSeek-inspired architecture. Each model is trained for 600B tokens. Our results establish a surprising robustness to distribution shifts for MoEs using both Sinkhorn-Balanced and Z-and-Aux-loss-balanced routing algorithms, even in MoEs continually pre-trained without replay. Moreover, we show that MoE LLMs maintain their sample efficiency (relative to a FLOP-matched dense model) during CPT and that they can match the performance of a fully re-trained MoE at a fraction of the cost.
2025-09-26
TMLR (accepté)
openreview.net
openreview.net
Investigating Faithfulness in Large Audio Language Models
Lovenya Jain
Pooneh Mousavi
Mirco Ravanelli
Cem Subakan
Faithfulness measures whether chain-of-thought (CoT) representations accurately reflect a model's decision process and can be used as reliab… (voir plus)le explanations. Prior work has shown that CoTs from text-based LLMs are often unfaithful. This question has not been explored for large audio-language models (LALMs), where faithfulness is critical for safety-sensitive applications. Reasoning in LALMs is also more challenging, as models must first extract relevant clues from audio before reasoning over them. In this paper, we investigate the faithfulness of CoTs produced by several LALMs by applying targeted interventions, including paraphrasing, filler token injection, early answering, and introducing mistakes, on two challenging reasoning datasets: SAKURA and MMAR. After going through the aforementioned interventions across several datasets and tasks, our experiments suggest that, LALMs generally produce CoTs that appear to be faithful to their underlying decision processes.
2025-09-26
ArXiv (prépublication)
arxiv.org
arxiv.org
TRUST: Test-Time Refinement using Uncertainty-Guided SSM Traverses
Sahar Dastani
Ali Bahri
Gustavo Adolfo Vargas Hakim
Moslem Yazdanpanah
Mehrdad Noori
David Osowiechi
Samuel Barbeau
Ismail Ben Ayed
Hervé Lombaert
Christian Desrosiers
State Space Models (SSMs) have emerged as efficient alternatives to Vision Transformers (ViTs), with VMamba standing out as a pioneering arc… (voir plus)hitecture designed for vision tasks. However, their generalization performance degrades significantly under distribution shifts. To address this limitation, we propose TRUST (Test-Time Refinement using Uncertainty-Guided SSM Traverses), a novel test-time adaptation (TTA) method that leverages diverse traversal permutations to generate multiple causal perspectives of the input image. Model predictions serve as pseudo-labels to guide updates of the Mamba-specific parameters, and the adapted weights are averaged to integrate the learned information across traversal scans. Altogether, TRUST is the first approach that explicitly leverages the unique architectural properties of SSMs for adaptation. Experiments on seven benchmarks show that TRUST consistently improves robustness and outperforms existing TTA methods.
2025-09-26
ArXiv (prépublication)
arxiv.org
arxiv.org
Acute respiratory distress syndrome in patients with cancer: the YELENNA prospective multinational observational cohort study.
Peter Schellongowski
Michael Darmon
Philipp Eller
Laveena Munshi
Tobias Liebregts
Victoria Metaxa
Luca Montini
Tobias Lahmer
F. Taccone
Andry Van de Louw
Martin Balik
P. Pickkers
Pleun Hemelaar
Hemang Yadav
Andreas Barratt-Due
T. Karvunidis
Jordi Riera
G. Martucci
I. Martín-Loeches
Pedro Castro … (voir 6 de plus)
Nina Buchtele
Virginie Lemiale
Stefan Hatzl
Guillaume Dumas
T. Staudinger
Elie Azoulay
2025-09-25
Intensive Care Medicine (publié)
doi.org
Acute respiratory distress syndrome in patients with cancer: the YELENNA prospective multinational observational cohort study.
Peter Schellongowski
Michael Darmon
Philipp Eller
Laveena Munshi
Tobias Liebregts
Victoria Metaxa
Luca Montini
Tobias Lahmer
F. Taccone
Andry Van de Louw
Martin Balik
P. Pickkers
Pleun Hemelaar
Hemang Yadav
Andreas Barratt-Due
T. Karvunidis
Jordi Riera
G. Martucci
I. Martín-Loeches
Pedro Castro … (voir 6 de plus)
Nina Buchtele
Virginie Lemiale
Stefan Hatzl
Guillaume Dumas
T. Staudinger
Elie Azoulay
2025-09-25
Intensive Care Medicine (publié)
doi.org
HEIST: A Graph Foundation Model for Spatial Transcriptomics and
Proteomics Data
Hiren Madhu
João Felipe Rocha
Tinglin Huang
Siddharth Viswanath
Smita Krishnaswamy
Rex Ying
2025-09-25
ArXiv (prépublication)
www.ncbi.nlm.nih.gov
Correction to: Assessing the adoption of security policies by developers in terraform across different cloud providers
Alexandre Verdet
Mohammad Hamdaqa
Leuson Da Silva
Foutse Khomh
2025-09-24
Empirical Software Engineering (publié)
doi.org
Correction to: Assessing the adoption of security policies by developers in terraform across different cloud providers
Alexandre Verdet
Mohammad Hamdaqa
Leuson Da Silva
Foutse Khomh
2025-09-24
Empirical Software Engineering (publié)
doi.org
Large Pre-Trained Models for Bimanual Manipulation in 3D
Hanna Yurchyk
Wei-Di Chang
Gregory Dudek
David Meger
We investigate the integration of attention maps from a pre-trained Vision Transformer into voxel representations to enhance bimanual roboti… (voir plus)c manipulation. Specifically, we extract attention maps from DINOv2, a self-supervised ViT model, and interpret them as pixel-level saliency scores over RGB images. These maps are lifted into a 3D voxel grid, resulting in voxel-level semantic cues that are incorporated into a behavior cloning policy. When integrated into a state-of-the-art voxel-based policy, our attention-guided featurization yields an average absolute improvement of 8.2% and a relative gain of 21.9% across all tasks in the RLBench bimanual benchmark.
2025-09-24
ArXiv (prépublication)
arxiv.org
arxiv.org