Agentic Reasoning on AndroidWorld
82.05Success RateEAPO-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| EAPO-8B2026.05 | 82.05 | |
| EAPO-4B2026.05 | 79.59 | |
| EAPO-1.7B/2B2026.05 | 76.36 | |
| UI-TARS-2-2509Category=Closed-source2026.05 | 73.38 | |
| Claude-4-Sonnet-0929Category=Closed-source2026.05 | 71.6 | |
| DeepMiner-Mano-72BCategory=Closed-source2026.05 | 68.48 | |
| Qwen3-VL-235B-A22BCategory=Open-source2026.05 | 62 | |
| DAPOCategory=Training Method2026.05 | 61.76 | |
| Seed1.5-VL-250717Category=Closed-source2026.05 | 61.75 | |
| LAMERCategory=Training Method2026.05 | 60.98 | |
| GiGPOCategory=Training Method2026.05 | 58.87 | |
| DART-GUI-7BCategory=Open-source2026.05 | 57.99 | |
| GRPOCategory=Training Method2026.05 | 55.48 | |
| OpenAI CUA o3Category=Closed-source2026.05 | 55.43 | |
| GUI-Owl-7BCategory=Open-source2026.05 | 52.04 | |
| OpenCUA-32BCategory=Open-source2026.05 | 51.66 | |
| ARPOCategory=Open-source2026.05 | 49.31 | |
| TianXi-Action-7BCategory=Closed-source2026.05 | 48.93 | |
| ZeroGUICategory=Open-source2026.05 | 47.52 | |
| UI-TARS-7BCategory=Open-source2026.05 | 33.04 |