Jiaming Wang: Blog Robotics, vision-language-action models and embodied AI https://jiaming.im/ 2026-07-29T00:00:00Z Jiaming Wang Robot Policies Need the Right Memory, Not an 8K Context Window https://jiaming.im/pages/blog/robot-memory.html 2026-07-29T00:00:00Z RoboTTT, WAM-TTT, and RoboSSM show that long context matters only when history becomes usable state—through task-relevant information, aligned meta-training, and memory matched to its timescale. Jiaming Wang How We Won IEEE ICRA Robotics Challenges Two Years in a Row https://jiaming.im/pages/blog/how-to-win.html 2026-07-26T00:00:00Z Lessons from winning ICRA navigation and manipulation challenges: start from the problem, decompose it from first principles, and test the assumptions that matter. Jiaming Wang OpenGEN-1: Reconstructing GEN-1 from First Principles https://jiaming.im/pages/blog/opengen.html 2026-07-26T00:00:00Z A first-principles reconstruction of GEN-1 as a physical foundation model trained on streams of interaction, with a proposed architecture, objective, data scale, and real-time inference system. Jiaming Wang VLA Is Not Dead—At Least Not Yet https://jiaming.im/pages/blog/wam.html 2026-07-26T00:00:00Z DreamZero and WAM split robot policy learning into future-video prediction and inverse dynamics—opening a powerful path beyond VLA, but at a steep deployment cost. Jiaming Wang Pi 0.6: Supervised Learning in RL's Clothing https://jiaming.im/pages/blog/pio06.html 2026-03-14T00:00:00Z How RECAP reveals that modern RL post-training is fundamentally supervised learning guided by value-based preference signals and environment interaction. Jiaming Wang CFGRL: The Theory Behind pi0.6* https://jiaming.im/pages/blog/cfgrl.html 2026-03-14T00:00:00Z How CFGRL reframes RL policy improvement as optimality-conditioned inference, then implements it via classifier-free guidance at sampling time. Jiaming Wang