215 citations · 811 across the 24 of their papers we have counts for
26 papers
Destruction is a General Strategy to Learn Generation; Diffusion's Strength is to Take it Seriously; Exploration is the Future
Pierre-André Noël
I present diffusion models as part of a family of machine learning techniques that withhold information from a model's input and train it to guess the withheld information. I argue…
DiffuMamba: High-Throughput Diffusion LMs with Mamba Backbone
Vaibhav Singh, Oleksiy Ostapenko, Pierre-André Noël +2
Diffusion language models (DLMs) have emerged as a promising alternative to autoregressive (AR) generation, yet their reliance on Transformer backbones limits inference efficiency…
Adaptive Diffusion Denoised Smoothing : Certified Robustness via Randomized Smoothing with Differentially Private Guided Denoising Diffusion
Frederick Shpilevskiy, Saiyue Lyu, Krishnamurthy Dj Dvijotham +2
We propose Adaptive Diffusion Denoised Smoothing, a method for certifying the predictions of a vision model against adversarial examples, while adapting to the input. Our key insig…
Unifying Autoregressive and Diffusion-Based Sequence Generation
Nima Fathi, Torsten Scholak, Pierre-André Noël
We present significant extensions to diffusion-based sequence generation models, blurring the line with autoregressive language models. We introduce hyperschedules, which assign di…
AlignVLM: Bridging Vision and Language Latent Spaces for Multimodal Document Understanding
Ahmed Masry, Juan A. Rodriguez, Tianyu Zhang +19
Aligning visual features with language embeddings is a key challenge in vision-language models (VLMs). The performance of such models hinges on having a good connector that maps vi…
RepLiQA: A Question-Answering Dataset for Benchmarking LLMs on Unseen Reference Content
Joao Monteiro, Pierre-Andre Noel, Etienne Marcotte +6
Large Language Models (LLMs) are trained on vast amounts of data, most of which is automatically scraped from the internet. This data includes encyclopedic documents that harbor a…
XC-Cache: Cross-Attending to Cached Context for Efficient LLM Inference
João Monteiro, Étienne Marcotte, Pierre-André Noël +5
In-context learning (ICL) approaches typically leverage prompting to condition decoder-only language model generation on reference information. Just-in-time processing of a context…
Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection
Charles Guille-Escuret, Pierre-André Noël, Ioannis Mitliagkas +2
Improving the reliability of deployed machine learning systems often involves developing methods to detect out-of-distribution (OOD) inputs. However, existing research often narrow…
Flaky Performances when Pretraining on Relational Databases
Shengchao Liu, David Vazquez, Jian Tang +1
We explore the downstream task performances for graph neural network (GNN) self-supervised learning (SSL) methods trained on subgraphs extracted from relational databases (RDBs). I…
On the Value of ML Models
Fabio Casati, Pierre-André Noël, Jie Yang
We argue that, when establishing and benchmarking Machine Learning (ML) models, the research community should favour evaluation metrics that better capture the value delivered by t…
Self-Organization of Dragon Kings
Yuansheng Lin, Keith Burghardt, Martin Rohden +2
The mechanisms underlying cascading failures are often modeled via the paradigm of self-organized criticality. Here we introduce a simple model where nodes self-organize to be eith…
Strategic tradeoffs in competitor dynamics on adaptive networks
Laurent Hébert-Dufresne, Antoine Allard, Pierre-André Noël +2
Recent empirical work highlights the heterogeneity of social competitions such as political campaigns: proponents of some ideologies seek debate and conversation, others create ech…
Quantifying dynamical spillover in co-evolving multiplex networks
Vikram S. Vijayaraghavan, Pierre-André Noël, Zeev Maoz +1
Multiplex networks (a system of multiple networks that have different types of links but share a common set of nodes) arise naturally in a wide spectrum of fields. Theoretical stud…
A Bottom-Up Model of Self-Organized Criticality on Networks
Pierre-André Noël, Charles D. Brummitt, Raissa M. D'Souza
The Bak-Tang-Wiesenfeld (BTW) sandpile process is an archetypal, stylized model of complex systems with a critical point as an attractor of their dynamics. This phenomenon, called…
Controlling Self-Organizing Dynamics on Networks Using Models that Self-Organize
Pierre-André Noël, Charles D. Brummitt, Raissa M. D'Souza
Controlling self-organizing systems is challenging because the system responds to the controller. Here we develop a model that captures the essential self-organizing mechanisms of…
Bond percolation on a class of correlated and clustered random graphs
Antoine Allard, Laurent Hébert-Dufresne, Pierre-André Noël +2
We introduce a formalism for computing bond percolation properties of a class of correlated and clustered random graphs. This class of graphs is a generalization of the Configurati…
Exact solution of bond percolation on small arbitrary graphs
Antoine Allard, Laurent Hébert-Dufresne, Pierre-André Noël +2
We introduce a set of iterative equations that exactly solves the size distribution of components on small arbitrary graphs after the random removal of edges. We also demonstrate h…
Epidemics on contact networks: a general stochastic approach
Pierre-André Noël, Antoine Allard, Laurent Hébert-Dufresne +2
Dynamics on networks is considered from the perspective of Markov stochastic processes. We partially describe the state of the system through network motifs and infer any missing d…
Structural preferential attachment: Stochastic process for the growth of scale-free, modular and self-similar systems
Laurent Hébert-Dufresne, Antoine Allard, Vincent Marceau +2
Many complex systems have been shown to share universal properties of organization, such as scale independence, modularity and self-similarity. We borrow tools from statistical phy…
Structural preferential attachment: Network organization beyond the link
Laurent Hébert-Dufresne, Antoine Allard, Vincent Marceau +2
We introduce a mechanism which models the emergence of the universal properties of complex networks, such as scale independence, modularity and self-similarity, and unifies them un…
Modeling the dynamical interaction between epidemics on overlay networks
Vincent Marceau, Pierre-André Noël, Laurent Hébert-Dufresne +2
Epidemics seldom occur as isolated phenomena. Typically, two or more viral agents spread within the same host population and may interact dynamically with each other. We present a…
Propagation on networks: an exact alternative perspective
Pierre-André Noël, Antoine Allard, Laurent Hébert-Dufresne +2
By generating the specifics of a network structure only when needed (on-the-fly), we derive a simple stochastic process that exactly models the time evolution of susceptible-infect…
Propagation dynamics on networks featuring complex topologies
Laurent Hébert-Dufresne, Pierre-André Noël, Vincent Marceau +2
Analytical description of propagation phenomena on random networks has flourished in recent years, yet more complex systems have mainly been studied through numerical means. In thi…
Adaptive networks: coevolution of disease and topology
Vincent Marceau, Pierre-André Noël, Laurent Hébert-Dufresne +2
Adaptive networks have been recently introduced in the context of disease propagation on complex networks. They account for the mutual interaction between the network topology and…