Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Few-shot Robot Manipulation on REASSEMBLE (6 held-out tasks)
Loading...
82
Success Rate
π0.5 fine-tuned (UB)
0.88
21.94
43
64.06
May 29, 2026
Success Rate
Updated 2d ago
Evaluation Results
Method
Method
Links
Success Rate
π0.5 fine-tuned (UB)
m=50, architecture=π0....
2026.05
82
π0.5-primitive
m=10, strategy=primiti...
2026.05
81
OpenVLA fine-tuned (UB)
m=50, architecture=Ope...
2026.05
79
OpenVLA-primitive
m=10, strategy=primiti...
2026.05
78
π0.5-primitive
m=5, strategy=primitiv...
2026.05
74
OpenVLA-primitive
m=5, strategy=primitiv...
2026.05
71
π0.5-primitive
m=3, strategy=primitiv...
2026.05
66
OpenVLA-primitive
m=3, strategy=primitiv...
2026.05
62
OpenVLA-flat
m=10, strategy=flat, a...
2026.05
61
π0.5-flat
m=10, strategy=flat, a...
2026.05
58
π0.5-primitive
m=1, strategy=primitiv...
2026.05
44
OpenVLA-flat
m=5, strategy=flat, ar...
2026.05
42
OpenVLA-primitive
m=1, strategy=primitiv...
2026.05
41
π0.5-flat
m=5, strategy=flat, ar...
2026.05
39
OpenVLA-flat
m=3, strategy=flat, ar...
2026.05
34
π0.5-flat
m=3, strategy=flat, ar...
2026.05
31
π0.5-primitive
m=0, strategy=primitiv...
2026.05
31
OpenVLA-primitive
m=0, strategy=primitiv...
2026.05
27
OpenVLA-flat
m=1, strategy=flat, ar...
2026.05
24
π0.5-flat
m=1, strategy=flat, ar...
2026.05
22
OpenVLA-flat
m=0, strategy=flat, ar...
2026.05
18
π0.5-flat
m=0, strategy=flat, ar...
2026.05
15
External-planner-only baseline
2026.05
4
Feedback
Search any
task
Search any
task