AI RESEARCH
Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisons
arXiv CS.LG
•
ArXi:2603.02115v2 Announce Type: replace-cross General-purpose robot reward models are typically trained to predict absolute task progress from expert nstrations, providing only local, frame-level supervision. While effective for expert nstrations, this paradigm scales poorly to large-scale robotics datasets where failed and suboptimal trajectories are abundant and assigning dense progress labels is ambiguous. We