Junwei Quan | LEAF Laboratoryhttp://leaf.utias.utoronto.ca/author/junwei-quan/Junwei QuanHugo Blox Builder (https://hugoblox.com)en-usWed, 07 Oct 2026 00:00:00 +0000http://leaf.utias.utoronto.ca/media/logo_hue87611c63d2e2a82e78083b2dc52db80_63556_300x300_fit_lanczos_3.pngJunwei Quanhttp://leaf.utias.utoronto.ca/author/junwei-quan/World-Model Policy Arbiter for Goal-Conditioned Reinforcement Learninghttp://leaf.utias.utoronto.ca/publication/quan-2026-wmpa/Wed, 07 Oct 2026 00:00:00 +0000http://leaf.utias.utoronto.ca/publication/quan-2026-wmpa/<h2 id="summary">Summary</h2> <p>WMPA selects among frozen goal-conditioned policies by simulating their futures in a learned world model and scoring them with a shared value function, without retraining the policies.</p>