Agentic Task Success on Textworld
75Success RateALMA
Evaluation Results
| Method | Links | |
|---|---|---|
| ALMAFoundation Model=GPT-5-mini2026.02 | 75 | |
| G-MemoryFoundation Model=GPT-5-mini2026.02 | 68.8 | |
| Trajectory RetrievalFoundation Model=GPT-5-mini2026.02 | 67 | |
| No MemoryFoundation Model=GPT-5-mini2026.02 | 60.5 | |
| Dynamic CheatsheetFoundation Model=GPT-5-mini2026.02 | 57.8 | |
| Reasoning BankFoundation Model=GPT-5-mini2026.02 | 56.1 | |
| ALMAFoundation Model=GPT-5-nano2026.02 | 6.2 | |
| No MemoryFoundation Model=GPT-5-nano2026.02 | 5.4 | |
| Reasoning BankFoundation Model=GPT-5-nano2026.02 | 5.3 | |
| Dynamic CheatsheetFoundation Model=GPT-5-nano2026.02 | 4.3 | |
| Trajectory RetrievalFoundation Model=GPT-5-nano2026.02 | 2.7 | |
| G-MemoryFoundation Model=GPT-5-nano2026.02 | 2.1 |