Mobile GUI Agents on AndroidWorld 138 tasks (test)
71.1Success RateUI-Mem-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| UI-Mem-8BParams.=8B, Inference-time memory retrieval=true2026.02 | 71.1 | |
| Seed1.8Params.=-, Inference-time memory retrieval=false2026.02 | 70.7 | |
| Gemini-2.5-ProParams.=-, Inference-time memory retrieval=false2026.02 | 69.7 | |
| Step-GUI-8BParams.=8B, Inference-time memory retrieval=false2026.02 | 67.7 | |
| UI-Mem-8BParams.=8B, Inference-time memory retrieval=false2026.02 | 66.8 | |
| GUI-Owl-7BParams.=7B, Inference-time memory retrieval=false2026.02 | 66.4 | |
| UI-Tars-1.5Params.=-, Inference-time memory retrieval=false2026.02 | 64.2 | |
| UI-Mem-4BParams.=4B, Inference-time memory retrieval=true2026.02 | 62.5 | |
| Seed1.5-VLParams.=-, Inference-time memory retrieval=false2026.02 | 62.1 | |
| UI-Mem-4BParams.=4B, Inference-time memory retrieval=false2026.02 | 58.2 | |
| MAI-UI-2BParams.=2B, Inference-time memory retrieval=false2026.02 | 49.1 | |
| UI-Venus-7BParams.=7B, Inference-time memory retrieval=false2026.02 | 49.1 | |
| Qwen3-VL-8BParams.=8B, Inference-time memory retrieval=false2026.02 | 47.6 | |
| Qwen3-VL-4BParams.=4B, Inference-time memory retrieval=false2026.02 | 45.3 | |
| Qwen3-VL-2BParams.=2B, Inference-time memory retrieval=false2026.02 | 36.4 | |
| UI-Tars-1.5-7BParams.=7B, Inference-time memory retrieval=false2026.02 | 30 | |
| Ferret-UI Lite-3BParams.=3B, Inference-time memory retrieval=false2026.02 | 28 | |
| ScaleCUA-3BParams.=3B, Inference-time memory retrieval=false2026.02 | 23.7 |