r5_c2
done · peak unseen_mismatch 89.8
α schedule
α 2.1319 ×25 α 2.1328 ×25 α 2.1351 ×25 α 2.139 ×25 α 2.1446 ×25 α 2.1522 ×25 α 2.1618 ×25 α 2.1736 ×25 α 2.1877 ×25 α 2.2041 ×25 α 2.223 ×25 α 2.2445 ×25 α 2.2685 ×25 α 2.2952 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 2.2952
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 | 50.8 ± 1.5 | 50.1 ± 1.8 | 49.9 ± 1.5 | 49.5 ± 1.9 |
| 50 | 51.0 ± 1.4 | 50.3 ± 1.6 | 49.9 ± 1.1 | 49.8 ± 1.4 |
| 75 | 51.1 ± 1.7 | 49.7 ± 1.8 | 50.2 ± 1.4 | 49.6 ± 1.4 |
| 100 | 51.7 ± 3.3 | 50.5 ± 1.6 | 50.6 ± 2.0 | 49.9 ± 1.7 |
| 125 | 56.4 ± 7.3 | 51.0 ± 2.9 | 53.8 ± 7.0 | 49.5 ± 2.2 |
| 150 | 65.6 ± 17.6 | 56.5 ± 9.1 | 62.8 ± 17.7 | 53.7 ± 8.1 |
| 175 | 74.8 ± 19.7 | 66.6 ± 15.3 | 72.5 ± 20.1 | 63.8 ± 14.2 |
| 200 | 82.3 ± 18.2 | 76.2 ± 17.2 | 79.2 ± 19.6 | 72.5 ± 17.7 |
| 225 | 89.0 ± 13.5 | 84.1 ± 13.4 | 85.3 ± 15.4 | 79.4 ± 14.4 |
| 250 | 95.3 ± 4.8 | 92.1 ± 6.3 | 91.2 ± 7.9 | 86.4 ± 8.7 |
| 275 | 97.2 ± 2.1 | 94.4 ± 4.6 | 93.7 ± 4.2 | 89.2 ± 5.9 |
| 300 | 97.6 ± 1.7 | 95.9 ± 2.1 | 93.6 ± 4.8 | 89.8 ± 6.1 |
| 325 | 97.1 ± 1.7 | 95.4 ± 2.5 | 91.4 ± 5.0 | 87.4 ± 6.8 |
| 350 | 97.1 ± 1.3 | 95.1 ± 1.8 | 90.4 ± 4.6 | 86.6 ± 5.9 |