RelayWAM

World-action model from Xinjiang University, Tongji University, CASIA and USTC that adds 34.96M-parameter transition layers to Fast-WAM so that the action pathway reads the backbone’s future-predictive features.

Xinjiang University et al. · Nov 17, 2026 · World-action · built on Fast-WAMReport an error
98.2%LIBERO · #36 of 83
72.2%LIBERO+ · #30 of 37
2Benchmarks reported
Nov 17, 2026Released

Benchmark results2

#30 of 3772.2%, self-reported by the authors, not a leaderboard entry
LIBEROSaturatedSimulation
#36 of 8398.2%, self-reported by the authors, not a leaderboard entry
Not reported on 13 benchmarks
Robustness drop: 98.2% on LIBERO → 72.2% on LIBERO-Plus. The perturbed variants show how much of the LIBERO score is memorization.
Wrong or missing number? Report an error