# Action Prediction on AITZ (test) Source: [Action Prediction on AITZ (test)]() Collection index: [llms.txt]() Data updated: 2026-03-09T09:00:34.069Z Evaluation of Action Prediction performance on the test split of the AITZ dataset, measured by metrics related to scrolling, clicking, and typing actions. 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 | Action Prediction | | Dataset (including any stated split) | AITZ (test) | | Metric | Scroll Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 17 of 17 | | 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 - [Scroll Accuracy](): Higher is better; unit: PERCENT. - [Click Accuracy (Type)](): Higher is better; unit: PERCENT. - [Click Accuracy (Match)](): Higher is better; unit: PERCENT. - [Type Accuracy (Type)](): Higher is better; unit: PERCENT. - [Type Accuracy (Match)](): Higher is better; unit: PERCENT. - [Press Accuracy](): Higher is better; unit: PERCENT. - [Stop Accuracy](): Higher is better; unit: PERCENT. - [Overall Accuracy (Type)](): Higher is better; unit: PERCENT. - [Overall Accuracy (Match)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Scroll Accuracy | Result date | Report paper | | --- | --- | --- | --- | | OS-Atlas | 76.12 | Not recorded | 69a519842d603e6de3e3c9b6 | | Auto-GUI | 61.4 | Not recorded | 69a519842d603e6de3e3c9b6 | | UGround | 58.22 | Not recorded | 69a519842d603e6de3e3c9b6 | | UITars | 56.5 | Not recorded | 69a519842d603e6de3e3c9b6 | | AriaUI | 53.73 | Not recorded | 69a519842d603e6de3e3c9b6 | | CoME | 52.07 | Not recorded | 69a519842d603e6de3e3c9b6 | | Qwen2VL-MoE | 48.75 | Not recorded | 69a519842d603e6de3e3c9b6 | | UI-S1 | 45.6 | Not recorded | 69a519842d603e6de3e3c9b6 | | Qwen2VL | 43.28 | Not recorded | 69a519842d603e6de3e3c9b6 | | UITars | 30.17 | Not recorded | 69a519842d603e6de3e3c9b6 | | GUI-R1 | 29.02 | Not recorded | 69a519842d603e6de3e3c9b6 | | MolmoE | 28.19 | Not recorded | 69a519842d603e6de3e3c9b6 | | GUI-R1 | 25.53 | Not recorded | 69a519842d603e6de3e3c9b6 | | ShowUI | 23.79 | Not recorded | 69a519842d603e6de3e3c9b6 | | Qwen2VL | 23.05 | Not recorded | 69a519842d603e6de3e3c9b6 | | DeepSeekVL2 | 17.94 | Not recorded | 69a519842d603e6de3e3c9b6 | | SeeClick | 11.14 | Not recorded | 69a519842d603e6de3e3c9b6 | ### OS-Atlas Scroll Accuracy: 76.12. Result ID: 69ae8c1b6dbb7075b956dd17. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"7B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### Auto-GUI Scroll Accuracy: 61.4. Result ID: 69ae8c1b6dbb7075b956dd0f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"700M","Source":"Reported from Zhang et al., 2024","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### UGround Scroll Accuracy: 58.22. Result ID: 69ae8c1b6dbb7075b956dd16. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"7B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### UITars Scroll Accuracy: 56.5. Result ID: 69ae8c1b6dbb7075b956dd18. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"7B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### AriaUI Scroll Accuracy: 53.73. Result ID: 69ae8c1b6dbb7075b956dd1d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"3.9B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### CoME Scroll Accuracy: 52.07. Result ID: 69ae8c1b6dbb7075b956dd1f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"5B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### Qwen2VL-MoE Scroll Accuracy: 48.75. Result ID: 69ae8c1b6dbb7075b956dd1c. [Reporting paper]() Recorded evaluation configuration: {"#Params":"3B","Finetuned with hybrid-capabilities reasoning":"true"}. This record does not assert independent reproduction. ### UI-S1 Scroll Accuracy: 45.6. Result ID: 69ae8c1b6dbb7075b956dd1a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"7B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### Qwen2VL Scroll Accuracy: 43.28. Result ID: 69ae8c1b6dbb7075b956dd15. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"7B","Finetuned with hybrid-capabilities reasoning":"true"}. This record does not assert independent reproduction. ### UITars Scroll Accuracy: 30.17. Result ID: 69ae8c1b6dbb7075b956dd12. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"2B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### GUI-R1 Scroll Accuracy: 29.02. Result ID: 69ae8c1b6dbb7075b956dd19. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"7B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### MolmoE Scroll Accuracy: 28.19. Result ID: 69ae8c1b6dbb7075b956dd1b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"1B","Finetuned with hybrid-capabilities reasoning":"true"}. This record does not assert independent reproduction. ### GUI-R1 Scroll Accuracy: 25.53. Result ID: 69ae8c1b6dbb7075b956dd13. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"3B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### ShowUI Scroll Accuracy: 23.79. Result ID: 69ae8c1b6dbb7075b956dd11. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"2B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. ### Qwen2VL Scroll Accuracy: 23.05. Result ID: 69ae8c1b6dbb7075b956dd10. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"2B","Finetuned with hybrid-capabilities reasoning":"true"}. This record does not assert independent reproduction. ### DeepSeekVL2 Scroll Accuracy: 17.94. Result ID: 69ae8c1b6dbb7075b956dd1e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#Params":"4.5B","Finetuned with hybrid-capabilities reasoning":"true"}. This record does not assert independent reproduction. ### SeeClick Scroll Accuracy: 11.14. Result ID: 69ae8c1b6dbb7075b956dd14. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"#Params":"7B","Finetuned with hybrid-capabilities reasoning":"false"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.