3D-LOTUS

Reported on 6 benchmarks across 1 task · 1 paper · 1 SOTA

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

Robots6 results

Robot ManipulationonRLBench
Succ. Rate (18 tasks, 100 demo/task)· 2024-10-02
83.1
best: 89.4 (EquAct)
SOTA
Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-guided 3D Policy arXiv:2410.01345
Robot ManipulationonRLBench
Inference Speed (fps)· 2024-10-02
9.5
best: 1.4 (EquAct)
Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-guided 3D Policy arXiv:2410.01345
Robot ManipulationonRLBench
Input Image Size· 2024-10-02
256
Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-guided 3D Policy arXiv:2410.01345
Robot ManipulationonRLBench
Training Time (A100 x hour)· 2024-10-02
40
best: 936 (3D Diffuser Actor)
Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-guided 3D Policy arXiv:2410.01345
Robot ManipulationonRLBench
Training Time (V100 x 8 x day)· 2024-10-02
0.28
best: 16 (PerAct (Evaluated in RVT))
Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-guided 3D Policy arXiv:2410.01345
Robot ManipulationonGEMBench
Average Success Rate· 2024-10-02
45.7
best: 48 (3D-LOTUS++)
Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-guided 3D Policy arXiv:2410.01345