The RLxF: RL from World Feedback workshop at ICML 2026 is getting some extra attention on X after organizer Shaohua Song framed it there around a simple question: what happens if "the world itself becomes the signal" instead of relying on human feedback alone. In a separate X post, Chelsea Finn wrote that her talk will focus on moving beyond the "scalar reward bottleneck" for robotics and LLMs.