# Desktop automation on WindowsAgentArena (WAA) v1 (test) Source: [Desktop automation on WindowsAgentArena (WAA) v1 (test)]() Collection index: [llms.txt]() Data updated: 2026-04-24T08:08:07.409Z Evaluation of agent performance for general desktop automation tasks using the test split of WindowsAgentArena version 1, reporting scores across Office, Web, System, Code, Media, and Utility categories. 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 | Desktop automation | | Dataset (including any stated split) | WindowsAgentArena (WAA) v1 (test) | | Metric | Overall Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 13 of 13 | | 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 - [Overall Score](): Higher is better; unit: PERCENT. - [Office Score](): Higher is better; unit: PERCENT. - [Web Score](): Higher is better; unit: PERCENT. - [System Score](): Higher is better; unit: PERCENT. - [Code Score](): Higher is better; unit: PERCENT. - [Media Score](): Higher is better; unit: PERCENT. - [Utility Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | Overall Score | Result date | Report paper | | --- | --- | --- | --- | | VLAA-GUI | 61 | Not recorded | 69eaf92c55a3c32af89f8a00 | | VLAA-GUI | 60.4 | Not recorded | 69eaf92c55a3c32af89f8a00 | | Agent S3 | 56.6 | Not recorded | 69eaf92c55a3c32af89f8a00 | | VLAA-GUI - Loop Breaker | 55.8 | Not recorded | 69eaf92c55a3c32af89f8a00 | | Agent S3 | 54.1 | Not recorded | 69eaf92c55a3c32af89f8a00 | | VLAA-GUI - Search Agent | 53.9 | Not recorded | 69eaf92c55a3c32af89f8a00 | | VLAA-GUI - Loop Breaker | 52.6 | Not recorded | 69eaf92c55a3c32af89f8a00 | | VLAA-GUI - Completeness Verifier | 51.3 | Not recorded | 69eaf92c55a3c32af89f8a00 | | VLAA-GUI - Completeness Verifier | 51.3 | Not recorded | 69eaf92c55a3c32af89f8a00 | | GTA1-32B | 51.2 | Not recorded | 69eaf92c55a3c32af89f8a00 | | UI-TARS-2 | 50.6 | Not recorded | 69eaf92c55a3c32af89f8a00 | | VLAA-GUI - Search Agent | 49.4 | Not recorded | 69eaf92c55a3c32af89f8a00 | | Qwen3-VL-32B | 31.7 | Not recorded | 69eaf92c55a3c32af89f8a00 | ### VLAA-GUI Overall Score: 61. Result ID: 69eb24d647f34ce924d3365a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Max Steps":"100","Backbone":"Gemini 3 Flash"}. This record does not assert independent reproduction. ### VLAA-GUI Overall Score: 60.4. Result ID: 69eb24d647f34ce924d33654. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Max Steps":"50","Backbone":"Gemini 3 Flash"}. This record does not assert independent reproduction. ### Agent S3 Overall Score: 56.6. Result ID: 69eb24d647f34ce924d33659. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Max Steps":"100","Backbone":"GPT-5"}. This record does not assert independent reproduction. ### VLAA-GUI - Loop Breaker Overall Score: 55.8. Result ID: 69eb24d647f34ce924d3365c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Max Steps":"100","Backbone":"Gemini 3 Flash","Ablation":"Remove Loop Breaker"}. This record does not assert independent reproduction. ### Agent S3 Overall Score: 54.1. Result ID: 69eb24d647f34ce924d33653. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Max Steps":"50","Backbone":"GPT-5"}. This record does not assert independent reproduction. ### VLAA-GUI - Search Agent Overall Score: 53.9. Result ID: 69eb24d647f34ce924d3365d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Max Steps":"100","Backbone":"Gemini 3 Flash","Ablation":"Remove Search Agent"}. This record does not assert independent reproduction. ### VLAA-GUI - Loop Breaker Overall Score: 52.6. Result ID: 69eb24d647f34ce924d33656. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Max Steps":"50","Backbone":"Gemini 3 Flash","Ablation":"Remove Loop Breaker"}. This record does not assert independent reproduction. ### VLAA-GUI - Completeness Verifier Overall Score: 51.3. Result ID: 69eb24d647f34ce924d33655. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Max Steps":"50","Backbone":"Gemini 3 Flash","Ablation":"Remove Completeness Verifier"}. This record does not assert independent reproduction. ### VLAA-GUI - Completeness Verifier Overall Score: 51.3. Result ID: 69eb24d647f34ce924d3365b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Max Steps":"100","Backbone":"Gemini 3 Flash","Ablation":"Remove Completeness Verifier"}. This record does not assert independent reproduction. ### GTA1-32B Overall Score: 51.2. Result ID: 69eb24d647f34ce924d33658. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Max Steps":"100","Backbone":"o3"}. This record does not assert independent reproduction. ### UI-TARS-2 Overall Score: 50.6. Result ID: 69eb24d647f34ce924d33652. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Max Steps":"50"}. This record does not assert independent reproduction. ### VLAA-GUI - Search Agent Overall Score: 49.4. Result ID: 69eb24d647f34ce924d33657. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Max Steps":"50","Backbone":"Gemini 3 Flash","Ablation":"Remove Search Agent"}. This record does not assert independent reproduction. ### Qwen3-VL-32B Overall Score: 31.7. Result ID: 69eb24d647f34ce924d33651. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Max Steps":"50"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.