Task Goal Completion on AppWorld Medium 48 tasks (test_normal)
43.8AccuracyNo compression
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| No compression2026.06 | 43.8 | 8.89 | 0 | |
| LRE2026.06 | 37.5 | 6.87 | 0 | |
| FIFOstrategy=recency2026.06 | 27.1 | 5.2 | 0 | |
| ACONmode=in-fork2026.06 | 16.7 | 4.92 | 5.62 |