Elaine Lau

I am a Machine Learning Research Engineer at Scale AI, building reasoning datasets, training pipelines, and safety evaluation frameworks for LLM-based browser agents. I partner with customers and research groups to deliver applied ML systems.

Previously, I was a researcher at Valence Labs – Recursion, developing QGFN by combining GFlowNets with action-value guidance for diverse solution discovery in drug discovery, and a graduate researcher at McGill University and Mila advised by Doina Precup and Emmanuel Bengio.

news

Oct 2025

Paper Rubrics as Rewards (arXiv)

Reinforcement learning beyond verifiable domains. Co-authors: A. Gunjal, A. Wang, et al.

arXiv OpenReview Project
Jun 2025

Conference Browser Agents Jailbreaks accepted to ICLR 2025

Refusal-trained LLMs can still be jailbroken as web agents; benchmark + code.

Details Code
Jan 2025

Invited talk AAAI Web Agents

Talk on jailbreaks in LLM-powered browser agents.
Dec 2024

Paper QGFN accepted to NeurIPS 2024

Controllable greediness with action values for GFlowNets (with S. Z. Lu, L. Pan, D. Precup, E. Bengio).

Details Code
2024–25

Work Scale AI — Reasoning datasets & agent safety

Built large-scale reasoning datasets and training pipelines; created BrowserART (100 adversarial behaviors, 40 sites) for agent red-teaming; led multi-agent data-review pipeline for real-time code filtering.
Dec 2023

Workshop DGFN at NeurIPS 2023 WS

Double GFlowNets for improved mode coverage (with N. Vemgal, D. Precup, E. Bengio).
Jul 2023

Workshop Replay Buffers for Mode Discovery — ICML 2023 WS

Empirical study demonstrating replay helps mode discovery in GFlowNets.
Feb 2023

Paper Deep Conservative RL for Ventilation — AAAI 2023

Equal-contribution clinical RL for personalized ventilator settings.

PDF
May 2022

Program Started M.Sc. at Mila/McGill

Advised by Doina Precup & Emmanuel Bengio; focus on GFlowNets + RL.

Looking for specifics? See publications or projects for links and code.