Congratulations to Yogesh Kulkarni, a PhD student in the People and Robots Lab, on his paper’s acceptance to @CVPR 2026!
TL;DR: AVATAR makes RL for video work:
1️⃣ Replay hard rollouts instead of wasting them
2️⃣ Credit to tokens that matter — planning & synthesis
📄: arxiv.org/abs/2508.03100
