# Language-conditioned Long-horizon Robot Manipulation on CALVIN GPT-4 enriched (Split D)

Source: [Language-conditioned Long-horizon Robot Manipulation on CALVIN GPT-4 enriched (Split D)](<https://www.sota2.com/research/sota/language-conditioned-long-horizon-robot-manipulation-on-calvin-gpt-4-enriched-split-d>)

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

Data updated: 2026-02-26T06:05:36.195Z

Evaluation of language-conditioned, long-horizon robot manipulation capabilities using the GPT-4 enriched version of Split D from the CALVIN dataset. Performance is measured by sequential success rates (1/5 through 5/5 steps) and average trajectory length.

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 | Language-conditioned long-horizon robot manipulation |
| Dataset (including any stated split) | CALVIN GPT-4 enriched Long-horizon (Split D) |
| Metric | Success Rate (1/5) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 3 of 3 |
| 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

- [Success Rate (1/5)](<https://www.sota2.com/research/sota/language-conditioned-long-horizon-robot-manipulation-on-calvin-gpt-4-enriched-split-d.md?metric=Success%20Rate%20(1%2F5)>): Higher is better; unit: PERCENT.

- [Success Rate (2/5)](<https://www.sota2.com/research/sota/language-conditioned-long-horizon-robot-manipulation-on-calvin-gpt-4-enriched-split-d.md?metric=Success%20Rate%20(2%2F5)>): Higher is better; unit: PERCENT.

- [Success Rate (3/5)](<https://www.sota2.com/research/sota/language-conditioned-long-horizon-robot-manipulation-on-calvin-gpt-4-enriched-split-d.md?metric=Success%20Rate%20(3%2F5)>): Higher is better; unit: PERCENT.

- [Success Rate (4/5)](<https://www.sota2.com/research/sota/language-conditioned-long-horizon-robot-manipulation-on-calvin-gpt-4-enriched-split-d.md?metric=Success%20Rate%20(4%2F5)>): Higher is better; unit: PERCENT.

- [Success Rate (5/5)](<https://www.sota2.com/research/sota/language-conditioned-long-horizon-robot-manipulation-on-calvin-gpt-4-enriched-split-d.md?metric=Success%20Rate%20(5%2F5)>): Higher is better; unit: PERCENT.

- [Avg Trajectory Length](<https://www.sota2.com/research/sota/language-conditioned-long-horizon-robot-manipulation-on-calvin-gpt-4-enriched-split-d.md?metric=Avg%20Trajectory%20Length>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Success Rate (1/5) | Result date | Report paper |
| --- | --- | --- | --- |
| LCB | 73.6 | Not recorded | 68fa79ef1e209131cea8014f |
| 3DDA | 65.2 | Not recorded | 68fa79ef1e209131cea8014f |
| RF | 62 | Not recorded | 68fa79ef1e209131cea8014f |

### LCB

Success Rate (1/5): 73.6. Result ID: 698b66387a967eddb76e4173.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79ef1e209131cea8014f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79ef1e209131cea8014f.md>)

Recorded evaluation configuration: {"Full Name":"Latent Codes as Bridges"}. This record does not assert independent reproduction.

### 3DDA

Success Rate (1/5): 65.2. Result ID: 698b66387a967eddb76e4172.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79ef1e209131cea8014f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa744b1e209131cea60daf.md>)

[Implementation](<https://www.sota2.com/3d-diffuser-actor.github.io>)

Recorded evaluation configuration: {"Description":"Policy from 3D Diffuser Actor"}. This record does not assert independent reproduction.

### RF

Success Rate (1/5): 62. Result ID: 698b66387a967eddb76e4171.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79ef1e209131cea8014f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b5.md>)

Recorded evaluation configuration: {"Description":"Own training of RoboFlamingo","Training Split":"ABC Calvin split"}. This record does not assert independent reproduction.

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