UI Agent Interaction on AgentNetBench
80.2Average SRQwen-3-VL-72B
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen-3-VL-72BHistory=5, Drop=✗2026.05 | 80.2 | 83.6 | 68.3 | 51.9 | — | |
| Qwen-3-VL-72BHistory=3, Drop=✗2026.05 | 79.3 | 82.5 | 67.4 | 50.8 | — | |
| Qwen3-VL-30B-A3BHistory=5, Drop=✗2026.05 | 77.8 | 81.3 | 66 | 49.1 | — | |
| Qwen3-VL-30B-A3BHistory=3, Drop=✗2026.05 | 77 | 80.4 | 65.1 | 48.2 | — | |
| Qwen-3-VL-72BHistory=1, Drop=✗2026.05 | 76.9 | 79.2 | 64.5 | 48.1 | — | |
| Qwen3-VL-30B-A3BHistory=1, Drop=✗2026.05 | 74.1 | 76.5 | 61.3 | 44.6 | — | |
| ReVisionHistory=5, Drop=✓2026.05 | 73.8 | 77 | 61 | 44.2 | — | |
| ReVisionHistory=5, Drop=✗2026.05 | 72.5 | 76.2 | 59.8 | 42.9 | — | |
| OpenCUAHistory=5, Drop=✗2026.05 | 72.4 | 76.1 | 59.7 | 42.8 | — | |
| Qwen-3-VL-8BHistory=5, Drop=✗2026.05 | 72.2 | 75.2 | 59.3 | 43.4 | — | |
| OpenCUAHistory=3, Drop=✗2026.05 | 72.1 | 75.8 | 59.4 | 42.5 | — | |
| ReVisionHistory=3, Drop=✗2026.05 | 72 | 75.9 | 59.3 | 42.4 | — | |
| Qwen-3-VL-8BHistory=3, Drop=✗2026.05 | 71.9 | 74.6 | 58.8 | 43.1 | — | |
| ReVisionHistory=3, Drop=✓2026.05 | 70.7 | 74.4 | 58.1 | 41.7 | — | |
| UI-TARS-1.5-7BHistory=5, Drop=✗2026.05 | 68.9 | 72.3 | 54.5 | 40.6 | — | |
| Qwen-3-VL-8BHistory=1, Drop=✗2026.05 | 68.7 | 73.4 | 54.9 | 40.2 | — | |
| OpenCUAHistory=1, Drop=✗2026.05 | 68.4 | 72 | 55.6 | 39.8 | — | |
| UI-TARS-1.5-7BHistory=3, Drop=✗2026.05 | 68.4 | 71.8 | 53.8 | 40.2 | — | |
| UI-TARS-1.5-7BHistory=1, Drop=✗2026.05 | 66.9 | 70.4 | 52.3 | 38.8 | — | |
| UI-TARS-72B-DPOHistory=5, Drop=✗2026.05 | 63.8 | 67.8 | 49.6 | 36 | — | |
| UI-TARS-72B-DPOHistory=3, Drop=✗2026.05 | 63.2 | 67.1 | 49 | 35.5 | — | |
| UI-TARS-72B-DPOHistory=1, Drop=✗2026.05 | 62.3 | 66.2 | 48.1 | 34.7 | — | |
| Kimi-VL-A3BHistory=5, Drop=✗2026.05 | 50.9 | 55.2 | 40.4 | 28.1 | — | |
| Kimi-VL-A3BHistory=3, Drop=✗2026.05 | 49.8 | 54 | 39.5 | 27.2 | — | |
| Kimi-VL-A3BHistory=1, Drop=✗2026.05 | 47.2 | 51.6 | 37.8 | 25.9 | — | |
| Qwen2.5-VL-72BHistory=5, Drop=✗2026.05 | 41.5 | 44.4 | 32.4 | 21.7 | — | |
| Qwen2.5-VL-72BHistory=3, Drop=✗2026.05 | 40.8 | 43.6 | 31.8 | 21 | — | |
| Qwen2.5-VL-32BHistory=5, Drop=✗2026.05 | 38.7 | 41 | 30.5 | 20.2 | — | |
| Qwen2.5-VL-72BHistory=1, Drop=✗2026.05 | 38.7 | 42.1 | 30.5 | 20.2 | — | |
| Qwen2.5-VL-32BHistory=3, Drop=✗2026.05 | 37.9 | 40.2 | 29.8 | 19.5 | — | |
| Qwen2.5-VL-32BHistory=1, Drop=✗2026.05 | 35.6 | 38.5 | 28.2 | 18.4 | — | |
| AGMem (Ours)Backbone Model=GPT-5.4-mini2026.06 | 28.8 | — | — | — | 34.6 | |
| + visual memoryBackbone Model=GPT-5.4-mini2026.06 | 28.1 | — | — | — | 33.8 | |
| Qwen2.5-VL-7BHistory=5, Drop=✗2026.05 | 25.8 | 27 | 20.1 | 13.3 | — | |
| Vanilla agentBackbone Model=GPT-5.4-mini2026.06 | 25.8 | — | — | — | 24.2 | |
| Qwen2.5-VL-7BHistory=3, Drop=✗2026.05 | 25 | 26.4 | 19.5 | 12.7 | — | |
| Qwen2.5-VL-7BHistory=1, Drop=✗2026.05 | 23.4 | 25.1 | 18.3 | 11.9 | — |