LIBERO Benchmarks
The compute breakdown — approximately 78% gradient, 21% rollout — is consistent across all three LIBERO benchmarks tested. The evaluation is limited to single-arm, short-horizon tasks; longer-horizon and bimanual coordination scenarios are…
1 sources - 3 claims
The compute breakdown — approximately 78% gradient, 21% rollout — is consistent across all three LIBERO benchmarks tested. The evaluation is limited to single-arm, short-horizon tasks; longer-horizon and bimanual coordination scenarios are not tested. Three LIBERO suites are used for evaluation: LIBERO-Object for object-centric pick-and-place, LIBERO-Spatial for spatial-relation following, and LIBERO-Goal for goal-conditioned manipulation.