Mobile GUI Agent Navigation on MemGUI-Bench
42.97Success RateTask-State Representation (Kimi-k2.5)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Task-State Representation (Kimi-k2.5)Base Model=Kimi-k2.5, Context=Proposed2026.07 | 42.97 | 55.8 | 0.78 | |
| Kimi-k2.5Context=Baseline2026.07 | 42.19 | 51.6 | — | |
| Task-State Representation (Qwen3.5-plus)Base Model=Qwen3.5-plus, Context=Proposed2026.07 | 28.91 | 59.8 | 2.35 | |
| Qwen3.5-plusContext=Baseline2026.07 | 26.56 | 60.5 | — |