# Visual Grounding on ScreenSpot V2 (Device/Input Modality Breakdown)

Source: [Visual Grounding on ScreenSpot V2 (Device/Input Modality Breakdown)](<https://www.sota2.com/research/sota/visual-grounding-on-screenspot-v2-device-input-modality-breakdown>)

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

Data updated: 2026-04-23T08:28:14.704Z

Evaluation of Visual Grounding performance on the ScreenSpot V2 dataset, reporting accuracy broken down by device type (Mobile, Desktop, Web) and input modality (Text, Icon).

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 | Visual Grounding |
| Dataset (including any stated split) | ScreenSpot V2 |
| Metric | Average 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

- [Average Accuracy](<https://www.sota2.com/research/sota/visual-grounding-on-screenspot-v2-device-input-modality-breakdown.md?metric=Average%20Accuracy>): Higher is better; unit: PERCENT.

- [Accuracy (Mobile, Text)](<https://www.sota2.com/research/sota/visual-grounding-on-screenspot-v2-device-input-modality-breakdown.md?metric=Accuracy%20(Mobile%2C%20Text)>): Higher is better; unit: PERCENT.

- [Accuracy (Mobile, Icon)](<https://www.sota2.com/research/sota/visual-grounding-on-screenspot-v2-device-input-modality-breakdown.md?metric=Accuracy%20(Mobile%2C%20Icon)>): Higher is better; unit: PERCENT.

- [Accuracy (Desktop, Text)](<https://www.sota2.com/research/sota/visual-grounding-on-screenspot-v2-device-input-modality-breakdown.md?metric=Accuracy%20(Desktop%2C%20Text)>): Higher is better; unit: PERCENT.

- [Accuracy (Desktop, Icon)](<https://www.sota2.com/research/sota/visual-grounding-on-screenspot-v2-device-input-modality-breakdown.md?metric=Accuracy%20(Desktop%2C%20Icon)>): Higher is better; unit: PERCENT.

- [Accuracy (Web, Text)](<https://www.sota2.com/research/sota/visual-grounding-on-screenspot-v2-device-input-modality-breakdown.md?metric=Accuracy%20(Web%2C%20Text)>): Higher is better; unit: PERCENT.

- [Accuracy (Web, Icon)](<https://www.sota2.com/research/sota/visual-grounding-on-screenspot-v2-device-input-modality-breakdown.md?metric=Accuracy%20(Web%2C%20Icon)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Average Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| LPO-8B | 90.5 | Not recorded | 68fa79731e209131cea7db23 |
| UI-TARS-72B | 90.3 | Not recorded | 68fa79731e209131cea7db23 |
| SE-GUI-7B | 90.3 | Not recorded | 68fa79731e209131cea7db23 |
| Qwen2.5-VL-7B | 88.8 | Not recorded | 68fa79731e209131cea7db23 |
| UGround-V1-7B | 87.6 | Not recorded | 68fa79731e209131cea7db23 |
| Aguvis-7B | 86 | Not recorded | 68fa79731e209131cea7db23 |
| UI-TARS-2B | 84.7 | Not recorded | 68fa79731e209131cea7db23 |
| OS-Atlas-7B | 84.1 | Not recorded | 68fa79731e209131cea7db23 |
| Qwen2.5-VL-3B | 80.9 | Not recorded | 68fa79731e209131cea7db23 |
| GPT-4o + OmniParser-v2 | 80.7 | Not recorded | 68fa79731e209131cea7db23 |
| OS-Atlas-4B | 71.9 | Not recorded | 68fa79731e209131cea7db23 |
| Operator | 70.5 | Not recorded | 68fa79731e209131cea7db23 |
| Magma-8B | 61.5 | Not recorded | 68fa79731e209131cea7db23 |
| SeeClick-9.6B | 55.1 | Not recorded | 68fa79731e209131cea7db23 |

### LPO-8B

Average Accuracy: 90.5. Result ID: 69e9d81647f34ce924d0f8ec.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

Recorded evaluation configuration: {"Category":"GUI-specific (RL)"}. This record does not assert independent reproduction.

### UI-TARS-72B

Average Accuracy: 90.3. Result ID: 69e9d81647f34ce924d0f8ea.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76a61e209131cea6e2ba.md>)

Recorded evaluation configuration: {"Category":"GUI-specific (SFT)"}. This record does not assert independent reproduction.

### SE-GUI-7B

Average Accuracy: 90.3. Result ID: 69e9d81647f34ce924d0f8eb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

Recorded evaluation configuration: {"Category":"GUI-specific (RL)"}. This record does not assert independent reproduction.

### Qwen2.5-VL-7B

Average Accuracy: 88.8. Result ID: 69e9d81647f34ce924d0f8e2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

Recorded evaluation configuration: {"Category":"General open-source"}. This record does not assert independent reproduction.

### UGround-V1-7B

Average Accuracy: 87.6. Result ID: 69e9d81647f34ce924d0f8e9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

Recorded evaluation configuration: {"Category":"GUI-specific (SFT)"}. This record does not assert independent reproduction.

### Aguvis-7B

Average Accuracy: 86. Result ID: 69e9d81647f34ce924d0f8e8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

Recorded evaluation configuration: {"Category":"GUI-specific (SFT)"}. This record does not assert independent reproduction.

### UI-TARS-2B

Average Accuracy: 84.7. Result ID: 69e9d81647f34ce924d0f8e6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76a61e209131cea6e2ba.md>)

Recorded evaluation configuration: {"Category":"GUI-specific (SFT)"}. This record does not assert independent reproduction.

### OS-Atlas-7B

Average Accuracy: 84.1. Result ID: 69e9d81647f34ce924d0f8e7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75791e209131cea68351.md>)

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

Recorded evaluation configuration: {"Category":"GUI-specific (SFT)"}. This record does not assert independent reproduction.

### Qwen2.5-VL-3B

Average Accuracy: 80.9. Result ID: 69e9d81647f34ce924d0f8e1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

Recorded evaluation configuration: {"Category":"General open-source"}. This record does not assert independent reproduction.

### GPT-4o + OmniParser-v2

Average Accuracy: 80.7. Result ID: 69e9d81647f34ce924d0f8e0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa744e1e209131cea60edb.md>)

Recorded evaluation configuration: {"Category":"Proprietary"}. This record does not assert independent reproduction.

### OS-Atlas-4B

Average Accuracy: 71.9. Result ID: 69e9d81647f34ce924d0f8e5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75791e209131cea68351.md>)

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

Recorded evaluation configuration: {"Category":"GUI-specific (SFT)"}. This record does not assert independent reproduction.

### Operator

Average Accuracy: 70.5. Result ID: 69e9d81647f34ce924d0f8df.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

Recorded evaluation configuration: {"Category":"Proprietary"}. This record does not assert independent reproduction.

### Magma-8B

Average Accuracy: 61.5. Result ID: 69e9d81647f34ce924d0f8e4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

Recorded evaluation configuration: {"Category":"GUI-specific (SFT)"}. This record does not assert independent reproduction.

### SeeClick-9.6B

Average Accuracy: 55.1. Result ID: 69e9d81647f34ce924d0f8e3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79731e209131cea7db23.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72441e209131cea55627.md>)

[Implementation](<https://github.com/njucckevin/SeeClick>)

Recorded evaluation configuration: {"Category":"GUI-specific (SFT)"}. This record does not assert independent reproduction.

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