# Embodied Planning & Correction on Cosmos

Source: [Embodied Planning & Correction on Cosmos](<https://www.sota2.com/research/sota/embodied-planning-and-correction-on-cosmos>)

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

Data updated: 2026-06-11T09:26:03.609Z

Evaluation of model performance on the Embodied Planning & Correction task using the Cosmos dataset, measured by Accuracy.

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 | Embodied Planning & Correction |
| Dataset (including any stated split) | Cosmos |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 14 of 14 |
| 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

- [Accuracy](<https://www.sota2.com/research/sota/embodied-planning-and-correction-on-cosmos.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Embodied-R1.5 | 69.3 | Not recorded | 6a2a40eed33dfb23e78391ab |
| GPT-5.4 | 69.2 | Not recorded | 6a2a40eed33dfb23e78391ab |
| Qwen3-VL-8B | 64 | Not recorded | 6a2a40eed33dfb23e78391ab |
| Gemini-2.5-Pro | 62.8 | Not recorded | 6a2a40eed33dfb23e78391ab |
| Pelican-VL-1.0 | 62.6 | Not recorded | 6a2a40eed33dfb23e78391ab |
| VeBrain | 58.8 | Not recorded | 6a2a40eed33dfb23e78391ab |
| Mimo-Embodied | 56.8 | Not recorded | 6a2a40eed33dfb23e78391ab |
| Embodied-R1 | 55.2 | Not recorded | 6a2a40eed33dfb23e78391ab |
| Gemini-Robotics-ER-1.5 | 53.7 | Not recorded | 6a2a40eed33dfb23e78391ab |
| GPT-4o | 53.3 | Not recorded | 6a2a40eed33dfb23e78391ab |
| InternVL3.5-8B | 48.2 | Not recorded | 6a2a40eed33dfb23e78391ab |
| RoboBrain-2.0-7B | 33.8 | Not recorded | 6a2a40eed33dfb23e78391ab |
| Magma | 32.8 | Not recorded | 6a2a40eed33dfb23e78391ab |
| Molmo-D-7B | 22.2 | Not recorded | 6a2a40eed33dfb23e78391ab |

### Embodied-R1.5

Accuracy: 69.3. Result ID: 6a2a7e8afbc373592c27ed28.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

[Implementation](<https://github.com/pickxiguapi/Embodied-R1.5>)

Recorded evaluation configuration: {"Model Type":"Ours"}. This record does not assert independent reproduction.

### GPT-5.4

Accuracy: 69.2. Result ID: 6a2a7e8afbc373592c27ed1c.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

Recorded evaluation configuration: {"Model Type":"General-Purpose VLM"}. This record does not assert independent reproduction.

### Qwen3-VL-8B

Accuracy: 64. Result ID: 6a2a7e8afbc373592c27ed1e.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

Recorded evaluation configuration: {"Model Type":"General-Purpose VLM"}. This record does not assert independent reproduction.

### Gemini-2.5-Pro

Accuracy: 62.8. Result ID: 6a2a7e8afbc373592c27ed1d.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

Recorded evaluation configuration: {"Model Type":"General-Purpose VLM"}. This record does not assert independent reproduction.

### Pelican-VL-1.0

Accuracy: 62.6. Result ID: 6a2a7e8afbc373592c27ed25.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/690988753ef7ccd74e5e5a1d.md>)

Recorded evaluation configuration: {"Model Type":"Embodied VLM"}. This record does not assert independent reproduction.

### VeBrain

Accuracy: 58.8. Result ID: 6a2a7e8afbc373592c27ed23.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78fa1e209131cea7b711.md>)

Recorded evaluation configuration: {"Model Type":"Embodied VLM"}. This record does not assert independent reproduction.

### Mimo-Embodied

Accuracy: 56.8. Result ID: 6a2a7e8afbc373592c27ed24.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/691ff204d6b0dd334292e641.md>)

[Implementation](<https://github.com/XiaomiMiMo/MiMo-Embodied>)

Recorded evaluation configuration: {"Model Type":"Embodied VLM"}. This record does not assert independent reproduction.

### Embodied-R1

Accuracy: 55.2. Result ID: 6a2a7e8afbc373592c27ed27.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

Recorded evaluation configuration: {"Model Type":"Embodied VLM"}. This record does not assert independent reproduction.

### Gemini-Robotics-ER-1.5

Accuracy: 53.7. Result ID: 6a2a7e8afbc373592c27ed21.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77a21e209131cea747ed.md>)

Recorded evaluation configuration: {"Model Type":"Embodied VLM"}. This record does not assert independent reproduction.

### GPT-4o

Accuracy: 53.3. Result ID: 6a2a7e8afbc373592c27ed1b.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

Recorded evaluation configuration: {"Model Type":"General-Purpose VLM"}. This record does not assert independent reproduction.

### InternVL3.5-8B

Accuracy: 48.2. Result ID: 6a2a7e8afbc373592c27ed1f.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af01e209131cea85dc9.md>)

Recorded evaluation configuration: {"Model Type":"General-Purpose VLM"}. This record does not assert independent reproduction.

### RoboBrain-2.0-7B

Accuracy: 33.8. Result ID: 6a2a7e8afbc373592c27ed22.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77931e209131cea74136.md>)

Recorded evaluation configuration: {"Model Type":"Embodied VLM"}. This record does not assert independent reproduction.

### Magma

Accuracy: 32.8. Result ID: 6a2a7e8afbc373592c27ed26.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa770b1e209131cea70b91.md>)

[Implementation](<https://microsoft.github.io/Magma>)

Recorded evaluation configuration: {"Model Type":"Embodied VLM"}. This record does not assert independent reproduction.

### Molmo-D-7B

Accuracy: 22.2. Result ID: 6a2a7e8afbc373592c27ed20.

[Reporting paper](<https://www.sota2.com/research/paper/6a2a40eed33dfb23e78391ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76101e209131cea6b37c.md>)

[Implementation](<https://molmo.allenai.org/>)

Recorded evaluation configuration: {"Model Type":"General-Purpose VLM"}. This record does not assert independent reproduction.

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