Event

PhysLadder: A Hierarchical Simulation-to-Real Contact-Rich Manipulation Benchmark

1 reports · 1 sources

Haibao Yu and other authors propose and release PhysLadder, a hierarchical simulation-to-real benchmark for evaluating physical interaction capabilities in contact-intensive manipulation. It contains 16 household tasks, divided into three cumulative levels: contact selection and geometric alignment (L1), force regulation under sustained contact (L2), and coordinated contact reconfiguration and force regulation (L3). Each task is instantiated in both physically and visually matched simulation and real environments, sharing task definitions and success criteria. The authors used this benchmark to conduct a zero-shot evaluation of GPT-6 Astra, with three trials per task in each environment.