Buckets:
Full 16-task results — AAM checkpoint-60000 (our harness, 10 eps/task)
| Suite | Task | Success |
|---|---|---|
| Counting | BinFill | 1/10 |
| Counting | PickXtimes | 0/10 |
| Counting | SwingXtimes | 1/10 |
| Counting | StopCube | 0/10 |
| Permanence | VideoUnmask | 2/10 |
| Permanence | VideoUnmaskSwap | 4/10 |
| Permanence | ButtonUnmask | 2/10 |
| Permanence | ButtonUnmaskSwap | 0/10 |
| Reference | PickHighlight | 2/10 |
| Reference | VideoRepick | 1/10 |
| Reference | VideoPlaceButton | 1/10 |
| Reference | VideoPlaceOrder | 1/10 |
| Imitation | MoveCube | 0/10 |
| Imitation | InsertPeg | 0/10 |
| Imitation | PatternLock | 1/10 |
| Imitation | RouteStick | 1/10 |
Suite totals: Counting 2/40 (5.0%) · Permanence 8/40 (20.0%) · Reference 5/40 (12.5%) · Imitation 2/40 (5.0%) · Total 17/160 = 10.6%
Baselines (official, 50 eps): vanilla 10.12 (C14/P9/R11/I6.5) · HAMLET 16.5 (C17.5/P19.5/R15/I14). Harness validated: reference HAMLET checkpoint scores Counting 6/40 on this harness ≈ official 17.5 (10-ep noise).
Xet Storage Details
- Size:
- 1.04 kB
- Xet hash:
- eee7750f24f5ea912d08af05eb94d8023fc15a9009798b57d25f20dbf74f2c71
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.