SAM2Act

Reported on 4 benchmarks across 1 task

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Robots4 results

Robot ManipulationonRLBench
Input Image Size
128
best: 256 (EquAct)
Robot ManipulationonRLBench
Succ. Rate (18 tasks, 100 demo/task)
86.8
best: 89.4 (EquAct)
Robot ManipulationonRLBench
Training Time (A100 x hour)
200
best: 936 (3D Diffuser Actor)
Robot ManipulationonThe COLOSSEUM
Average decrease average across all perturbations
-4.3