r6_c0
done · peak unseen_mismatch 72.1
α schedule
α 1.9838 ×25 α 2.7442 ×25 α 3.1925 ×25 α 2.564 ×25 α 2.7625 ×25 α 4.034 ×25 α 4.2392 ×25 α 4.4234 ×25 α 3.3371 ×25 α 3.4493 ×25 α 3.5535 ×25 α 5.0 ×25 α 5.0 ×25 α 3.8293 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 3.8293
seen_* probed at reference α = 1.4 · unseen_* = held-out pairings, uniform (α-independent) · 1000 pairs per condition
| iteration | seen_match | seen_mismatch | unseen_match | unseen_mismatch |
|---|---|---|---|---|
| 25 | 50.6 ± 1.3 | 50.2 ± 1.8 | 50.0 ± 1.5 | 49.5 ± 1.5 |
| 50 | 50.9 ± 1.5 | 50.4 ± 1.5 | 50.0 ± 1.5 | 50.0 ± 1.6 |
| 75 | 51.4 ± 1.7 | 49.8 ± 1.6 | 50.4 ± 1.5 | 49.4 ± 1.8 |
| 100 | 52.3 ± 4.2 | 50.9 ± 2.0 | 51.1 ± 2.4 | 50.2 ± 1.6 |
| 125 | 59.6 ± 10.3 | 54.7 ± 7.7 | 54.9 ± 8.8 | 51.1 ± 6.0 |
| 150 | 68.1 ± 13.6 | 64.0 ± 12.3 | 62.0 ± 11.6 | 58.2 ± 9.7 |
| 175 | 77.2 ± 12.6 | 74.5 ± 12.3 | 69.0 ± 13.2 | 65.6 ± 12.1 |
| 200 | 82.6 ± 8.5 | 80.5 ± 7.9 | 71.9 ± 11.5 | 68.0 ± 10.4 |
| 225 | 87.3 ± 5.6 | 85.0 ± 5.5 | 75.4 ± 10.0 | 72.1 ± 9.9 |
| 250 | 87.6 ± 4.5 | 85.9 ± 4.0 | 74.4 ± 8.6 | 71.8 ± 8.4 |
| 275 | 88.2 ± 3.9 | 86.4 ± 3.3 | 74.3 ± 7.7 | 71.2 ± 7.2 |
| 300 | 86.8 ± 3.4 | 85.2 ± 2.6 | 72.2 ± 6.5 | 69.2 ± 5.8 |
| 325 | 86.2 ± 3.1 | 84.7 ± 2.4 | 70.5 ± 5.8 | 68.0 ± 5.5 |
| 350 | 87.6 ± 2.9 | 85.8 ± 2.1 | 72.1 ± 5.5 | 69.5 ± 5.0 |