Waterloo's R2L Lab recruiting 1-2 PhD students in generalist agents research
hllo_wrld · x · 2026-10-06
The Reading to Learn (R2L) Lab at the University of Waterloo's Cheriton School of Computer Science is recruiting 1-2 PhD students. The PI is on-site at COLM through Friday and invites anyone interested in grad school or faculty jobs to reach out.
The lab builds generalist agents that read, reason, and act across digital and physical environments, organized around four threads:
- Agent building & evaluation: Spider 2 (enterprise text-to-SQL), OSWorld, Computer Agent Arena, OpenCUA, and SynQuE — its platforms serve as primary evaluations at OpenAI, Anthropic, Google, and Salesforce.
- Generalization & reasoning: RL/SFT complementary reasoning, and ASH on self-improvement from unlabeled long-horizon experience.
- Language as supervision: replacing hand-engineered rewards with instructions, critiques, and explanations (Language Feedback Models, Text2Reward).
More from Companies & People
- 91 insiders vote on the most insightful people in AI data and post-training; Daniel Chang tops the list — FinanceYF5 · 2026-10-06
- Hacktoberfest Open-Source AI Challenge Launches: $2,450 in Prizes Across 17 Winners — jonmarkgo · 2026-10-06
- Patrick Collison says the next Stripe won't come from an AI lab but from legacy vertical software — amasad · 2026-10-06
- Engineers' hardest lesson is economics: patching legacy code often beats rewriting — tekbog · 2026-10-06
- Polymarket Puts 62% Odds on OpenAI Announcing Another Millennium Prize Solution — Polymarket · 2026-10-06
- OpenAI to apologize again to Australia over Medicare hack as artists warn of copyright 'roadkill' — nordicinst · 2026-10-06