# Inference Efficiency on Real-world Robotic Setup inference 1.0 (test)

Source: [Inference Efficiency on Real-world Robotic Setup inference 1.0 (test)](<https://www.sota2.com/research/sota/inference-efficiency-on-real-world-robotic-setup-inference-1-0-test>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-17T10:20:52.576Z

Evaluation of inference efficiency metrics, specifically measuring the latency components of the observation encoder and action expert, derived from a real-world robotic setup environment (version 1.0 test split).

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Inference Efficiency |
| Dataset (including any stated split) | Real-world Robotic Setup inference 1.0 (test) |
| Metric | Action Expert Latency (ms) |
| Unit | TIME |
| Direction | Lower is better |
| Results loaded | 4 of 4 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Action Expert Latency (ms)](<https://www.sota2.com/research/sota/inference-efficiency-on-real-world-robotic-setup-inference-1-0-test.md?metric=Action%20Expert%20Latency%20(ms)>): Lower is better; unit: TIME.

- [Observation Encoder Latency (ms)](<https://www.sota2.com/research/sota/inference-efficiency-on-real-world-robotic-setup-inference-1-0-test.md?metric=Observation%20Encoder%20Latency%20(ms)>): Lower is better; unit: TIME.

## Recorded results

| Method | Action Expert Latency (ms) | Result date | Report paper |
| --- | --- | --- | --- |
| DP3 | 58.77 | Not recorded | 69b8e01a0636e5bf4afbeb2f |
| DP+V+M | 66.32 | Not recorded | 69b8e01a0636e5bf4afbeb2f |
| DP | 67.94 | Not recorded | 69b8e01a0636e5bf4afbeb2f |
| R3DP | 69.78 | Not recorded | 69b8e01a0636e5bf4afbeb2f |

### DP3

Action Expert Latency (ms): 58.77. Result ID: 69b92ad047f34ce924a2761b.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e01a0636e5bf4afbeb2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74f31e209131cea64c76.md>)

[Implementation](<https://github.com/YanjieZe/3D-Diffusion-Policy>)

Recorded evaluation configuration: {"Policy Architecture":"3D Diffusion Policy","Hardware":"RTX 4090"}. This record does not assert independent reproduction.

### DP+V+M

Action Expert Latency (ms): 66.32. Result ID: 69b92ad047f34ce924a2761c.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e01a0636e5bf4afbeb2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b8e01a0636e5bf4afbeb2f.md>)

Recorded evaluation configuration: {"Policy Architecture":"DP with VGGT and MVFF","Hardware":"RTX 4090"}. This record does not assert independent reproduction.

### DP

Action Expert Latency (ms): 67.94. Result ID: 69b92ad047f34ce924a2761a.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e01a0636e5bf4afbeb2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72841e209131cea56f68.md>)

Recorded evaluation configuration: {"Policy Architecture":"Diffusion Policy","Hardware":"RTX 4090"}. This record does not assert independent reproduction.

### R3DP

Action Expert Latency (ms): 69.78. Result ID: 69b92ad047f34ce924a2761d.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e01a0636e5bf4afbeb2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b8e01a0636e5bf4afbeb2f.md>)

Recorded evaluation configuration: {"Policy Architecture":"R3DP (Ours)","Hardware":"RTX 4090"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
