GUI Agent Task on MemGUI-Bench Medium
35.7Precision@1M3A
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| M3AMemory Type=Memory Agent2026.06 | 35.7 | 50 | 48.7 | |
| MemGUI-Agent-235BMemory Type=ConAct, Protocol=zero-shot, Backbone=Qwen3-VL-235B-Thinking2026.06 | 35.7 | 61.9 | 52.3 | |
| Qwen3-VL-235B-ThinkingMemory Type=Action-Thought2026.06 | 28.6 | 47.6 | 38.8 | |
| MemGUI-8B-SFTMemory Type=ConAct, Protocol=fine-tuned, Backbone=Qwen3-VL-8B-Instruct2026.06 | 23.8 | 33.3 | 36.5 | |
| Qwen3-VL-235B-InstructMemory Type=Action-Thought2026.06 | 21.4 | 40.5 | 28.8 | |
| Agent-S2Memory Type=Memory Agent2026.06 | 19 | 42.9 | 42.5 | |
| T3AMemory Type=Memory Agent2026.06 | 16.7 | 45.2 | — | |
| Qwen3-VL-32B-InstructMemory Type=Action-Thought2026.06 | 14.3 | 23.8 | 22.4 | |
| GUI-Owl-1.5-32B-InstructMemory Type=Action-Thought2026.06 | 7.1 | 14.3 | 21.3 | |
| Qwen3-VL-8B-InstructMemory Type=Action-Thought2026.06 | 4.8 | 14.3 | 16 | |
| Mobile-Agent-EMemory Type=Memory Agent2026.06 | 2.4 | 2.4 | 4.4 | |
| GUI-Owl-1.5-8B-InstructMemory Type=Action-Thought2026.06 | 2.4 | 2.4 | 16.5 | |
| Mobile-Agent-V2Memory Type=Memory Agent2026.06 | 0 | 0 | 0 | |
| SeeActMemory Type=Rule-based2026.06 | 0 | 2.4 | 0.5 | |
| AppAgentMemory Type=Action-Thought2026.06 | 0 | 2.4 | 0 | |
| GUI-Owl-7BMemory Type=Action-Thought2026.06 | 0 | 2.4 | 5 | |
| UI-TARS-1.5-7BMemory Type=Multi-turn2026.06 | 0 | 0 | 1.7 | |
| UI-Venus-7BMemory Type=Action-Thought2026.06 | 0 | 0 | 1.2 | |
| CogAgentMemory Type=No History2026.06 | 0 | 0 | 0 |