Agent Success Rate on MiniGrid full-view
99MTEGPT-5.2
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-5.2Base=GPT-5.2, Setting=Base2026.04 | 99 | 99 | 99 | 99 | |
| ReDActBase=Qwen3-80B, Setting=+ GPT-5.22026.04 | 93 | 95 | 95 | 93 | |
| ReDActBase=Qwen3-80B, Setting=+ Qwen3-235B2026.04 | 93 | 91.5 | 94 | 90 | |
| Qwen3-235BBase=Qwen3-235B, Setting=Base2026.04 | 92.5 | 92.5 | 92.5 | 92.5 | |
| Qwen3-80BBase=Qwen3-80B, Setting=Base2026.04 | 85.5 | 85.5 | 85.5 | 85.5 | |
| ReDActBase=Llama4, Setting=+ GPT-5.22026.04 | 37 | 32.5 | 42 | 38 | |
| ReDActBase=Llama4, Setting=+ Qwen3-235B2026.04 | 33 | 35.5 | 37 | 29.5 | |
| Llama4Base=Llama4, Setting=Base2026.04 | 22 | 22 | 22 | 22 |