Alexandre Lacoste
Research Lead
AI Frontier Research
Alexandre is a Research Scientist at ServiceNow Research. His main interests revolve around LLM agents for solving UI tasks. Other interests involve causality, probabilistic machine learning, meta-learning and more. In his free time, he explores how AI can help solve climate change issues.
Alexandre joined Element AI in early 2017 as its first Research Scientist. Prior to Element AI, he worked for 3 years at Google in the Research Group for building end-to-end question-answering systems using deep learning. This system is currently in use by Google’s search engine to answer some of the most complex questions.
He obtained his PhD in theoretical machine learning, working with Mario Marchand and Francois Laviolette, during which he developed bridges between Bayes and PAC-Bayes theories. He obtained his Master’s degree with Douglas Eck in the former MILA where he applied machine learning to music. Finally, he studied Physics during his undergrad.
Intérêts
- Climate Change
- Causality
Publications
How to Train Your LLM Web Agent: A Statistical Diagnosis.
Neural Information Processing Systems (NeurIPS),
2025.
GEOBench-VLM: Benchmarking Vision-Language Models for Geospatial Tasks.
International Conference on Computer Vision (ICCV),
2025.
DoomArena: A framework for Testing AI Agents Against Evolving Security Threats.
Conference on Language Modeling (COLM),
2025.
DoomArena: A framework for Testing AI Agents Against Evolving Security Threats.
Workshop at the International Conference of Machine Learning (ICML),
2025.
How to Train Your LLM Web Agent: A Statistical Diagnosis (Oral).
Workshop at the International Conference of Machine Learning (ICML),
2025.
Context is Key: A Benchmark for Forecasting with Essential Textual Information.
International Conference on Machine Learning (ICML),
2025.
InsightBench: Evaluating Business Analytics Agents Through Multi-Step Insight Generation.
International Conference of Learning Representations (ICLR),
2025.
EarthView: A Large Scale Remote Sensing Dataset for Self-Supervision.
Workshop at the Winter Conference on Applications of Computer Vision (WACV),
2025.
The BrowserGym Ecosystem for Web Agent Research.
Transactions on Machine Learning Research (TMLR),
2025.
AgentMerge: Enhancing Generalization in Fine-Tuned LLM Agents.
Workshop at the Neural Information Processing Systems (NeurIPS),
2024.
Context is Key: A Benchmark for Forecasting with Essential Textual Information.
Workshop at the Neural Information Processing Systems (NeurIPS),
2024.
Fine-Tuning Web Agents: It Works, But It's Trickier Than You Think.
Workshop at the Neural Information Processing Systems (NeurIPS),
2024.
WorkArena++: Towards Compositional Planning and Reasoning-based Common Knowledge Work Tasks.
NeurIPS Datasets and Benchmarks Track (NeurIPS Datasets),
2024.
Context is Key: A Benchmark for Forecasting with Essential Textual Information.
Foundation Models for Time Series,
2024.
Fine-Tuning Web Agents: It Works, But It's Trickier Than You Think.
NOW AI Conference (NOWAI),
2024.
An Ecosystem for Web Agents: WorkArena, BrowserGym, AgentLab and more.
Montreal AI Symposium (MAIS),
2024.
Context is Key: A Benchmark for Forecasting with Essential Textual Information.
Montreal AI Symposium (MAIS),
2024.
WorkArena: How Capable are Web Agents at Solving Common Knowledge Work Tasks?.
International Conference on Machine Learning (ICML),
2024.
WorkArena: How Capable are Web Agents at Solving Common Knowledge Work Tasks?.
Workshop at the International Conference of Learning Representation (ICLR),
2024.
Capture the Flag: Uncovering Data Insights with Large Language Models.
Workshop at the Neural Information Processing Systems (NeurIPS),
2023.
The Unsolved Challenges of LLMs in Open-Ended Web Tasks: A Case Study.
Workshop at the Neural Information Processing Systems (NeurIPS),
2023.
Mastering the Unsupervised Reinforcement Learning Benchmark from Pixels.
International Conference on Machine Learning (ICML),
2023.
GEO-Bench: Toward Foundation Models for Earth Monitoring.
NeurIPS Datasets and Benchmarks Track (NeurIPS Datasets),
2023.
Choreographer: Learning and Adapting Skills in Imagination.
International Conference of Learning Representations (ICLR),
2023.
Attention for Compositional Modularity.
Workshop at the Neural Information Processing Systems (NeurIPS),
2022.
A General Purpose Neural Architecture for Geospatial Systems.
Workshop at the Neural Information Processing Systems (NeurIPS),
2022.
Choreographer: Learning and Adapting Skills in Imagination .
Workshop at the Neural Information Processing Systems (NeurIPS),
2022.
Continual Learning with self-selecting specialized modules through expansion and pruning.
Montreal AI Symposium (MAIS),
2022.
Unsupervised Model-based Pre-training for Data-efficient Reinforcement Learning from Pixels.
Workshop at the International Conference on Machine Learning (ICML),
2022.
Deconfounding Dynamic Treatment Regimes.
SSC,
2022.
Can Active Learning Preemptively Mitigate Fairness Issues?.
Workshop at the International Conference on Learning Representations (ICLR),
2022.
Disentanglement via Mechanism Sparsity Regularization: A New Principle for Nonlinear ICA.
Causal Learning and Reasoning (CLeaR),
2022.
Typing assumptions improve identification in causal discovery - theory and algorithms.
Causal Learning and Reasoning (CLeaR),
2022.
Tackling Climate Change with Machine Learning.
ACM Computing Surveys,
2022.
Toward Foundation Models for Earth Monitoring: Proposal for a Climate Change Benchmark.
Workshop at the Neural Information Processing Systems (NeurIPS),
2021.
Typing assumptions improve identification in causal discovery - Report and comments on future directions.
Workshop at the Neural Information Processing Systems (NeurIPS),
2021.
Beyond Trivial Counterfactual Explanations with Diverse Valuable Explanations.
International Conference on Computer Vision (ICCV),
2021.
Seasonal Contrast: Unsupervised Pre-Training from Uncurated Remote Sensing Data.
International Conference on Computer Vision (ICCV),
2021.
Typing assumptions improve identification in causal discovery.
Workshop at the International Conference on Machine Learning (ICML),
2021.
Variational Causal Networks: Approximate Bayesian Inference over Causal Structures.
Workshop at the International Conference on Machine Learning (ICML),
2021.
Counting Cows: Tracking Illegal Cattle Ranching From High-Resolution Satellite Imagery.
Workshop at the Neural Information Processing Systems (NeurIPS),
2020.
Differentiable Causal Discovery from Interventional Data.
Conference on Neural Information Processing Systems (NeurIPS),
2020.
Online Fast Adaptation and Knowledge Accumulation: a New Approach to Continual Learning.
Conference on Neural Information Processing Systems (NeurIPS),
2020.
Synbols: Probing Learning Algorithms with Synthetic Datasets.
Conference on Neural Information Processing Systems (NeurIPS),
2020.
Embedding Propagation: Smoother Manifold for Few-Shot Classification.
European Conference on Computer Vision (ECCV),
2020.
Bayesian active learning for production, a systematic study and a reusable library.
Workshop at the International Conference on Machine Learning (ICML),
2020.
Estimating Carbon Emissions of Artificial Intelligence [Opinion].
IEEE Technology and Society Magazine (TSM),
2020.
Online Fast Adaptation and Knowledge Accumulation: a New Approach to Continual Learning.
Workshop at the Conference on Computer Vision and Pattern Recognition (CVPR),
2020.
Stochastic Neural Network with Kronecker Flow.
International Conference on Artificial Intelligence and Statistics (AISTATS),
2020.
Gradient-Based Neural DAG Learning with Interventions.
Workshop at the International Conference on Learning Representations (ICLR),
2020.
Tackling Climate Change with Machine Learning.
Workshop at the Neural Information Processing Systems (NeurIPS),
2019.
Quantifying the Carbon Emissions of Machine Learning.
ArXiv,
2019.
Hierarchical Importance Weighted Autoencoders.
International Conference on Machine Learning (ICML),
2019.
Stochastic Neural Network with Kronecker Flow.
Workshop at the International Conference on Machine Learning (ICML),
2019.
On Difficulties of Probability Distillation.
International Conference on Learning Representations (ICLR),
2019.
Improving Explorability in Variational Inference with Annealed Variational Objectives.
Conference on Neural Information Processing Systems (NeurIPS),
2018.
TADAM: Task dependent adaptive metric for improved few-shot learning.
Conference on Neural Information Processing Systems (NeurIPS),
2018.
Synbols: Probing Learning Algorithms with Synthetic Datasets.
Montreal AI Symposium (MAIS),
2018.
Neural Autoregressive Flows.
International Conference on Machine Learning (ICML),
2018.
Learning Heuristics for the TSP by Policy Gradient.
International Conference on the Integration of Constraint Programming, Artificial Intelligence, and Operations Research,
2018.
Bayesian Hypernetworks.
Workshop at the Neural Information Processing Systems (NeurIPS),
2017.
Deep Prior.
Workshop at the Neural Information Processing Systems (NeurIPS),
2017.
Accurate Supervised and Semi-Supervised Machine Reading for Long Documents.
Conference on Empirical Methods in Natural Language Processing (EMNLP),
2017.
Coarse-to-Fine Question Answering for Long Documents.
Annual Meeting of the Association for Computational Linguistics (ACL),
2016.