Agentic Task Success on Baba Is AI
38Success RateDynamic Cheatsheet
Evaluation Results
| Method | Links | |
|---|---|---|
| Dynamic CheatsheetFoundation Model=GPT-5-mini2026.02 | 38 | |
| ALMAFoundation Model=GPT-5-mini2026.02 | 33.3 | |
| Trajectory RetrievalFoundation Model=GPT-5-mini2026.02 | 30.9 | |
| G-MemoryFoundation Model=GPT-5-mini2026.02 | 26.2 | |
| No MemoryFoundation Model=GPT-5-mini2026.02 | 21.4 | |
| Reasoning BankFoundation Model=GPT-5-mini2026.02 | 21.4 | |
| Trajectory RetrievalFoundation Model=GPT-5-nano2026.02 | 19 | |
| ALMAFoundation Model=GPT-5-nano2026.02 | 19 | |
| G-MemoryFoundation Model=GPT-5-nano2026.02 | 14.3 | |
| No MemoryFoundation Model=GPT-5-nano2026.02 | 9.5 | |
| Reasoning BankFoundation Model=GPT-5-nano2026.02 | 9.5 | |
| Dynamic CheatsheetFoundation Model=GPT-5-nano2026.02 | 9.5 |