r1_c0
done · peak unseen_mismatch 89.1
α schedule
α 1.0746 ×25 α 1.0764 ×25 α 1.0834 ×25 α 1.0997 ×25 α 1.1292 ×25 α 1.1761 ×25 α 1.2447 ×25 α 1.3392 ×25 α 1.4639 ×25 α 1.6234 ×25 α 1.8221 ×25 α 2.0644 ×25 α 2.3549 ×25 α 2.6981 ×25 α 3.0988 ×25 α 3.5614 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 3.5614
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 | 50.4 ± 1.5 | 49.8 ± 1.8 | 50.0 ± 1.7 | 49.9 ± 1.7 |
| 50 | 50.6 ± 1.6 | 50.2 ± 1.5 | 50.1 ± 1.3 | 49.8 ± 1.7 |
| 75 | 50.5 ± 1.2 | 49.5 ± 1.8 | 50.1 ± 1.1 | 49.8 ± 1.5 |
| 100 | 50.2 ± 1.8 | 50.2 ± 1.4 | 49.8 ± 1.2 | 50.1 ± 1.5 |
| 125 | 51.1 ± 1.5 | 49.7 ± 1.6 | 50.4 ± 2.0 | 49.5 ± 2.0 |
| 150 | 50.6 ± 1.7 | 50.0 ± 1.5 | 49.8 ± 1.8 | 49.8 ± 1.6 |
| 175 | 54.8 ± 8.4 | 47.8 ± 4.3 | 54.5 ± 8.4 | 48.1 ± 4.6 |
| 200 | 62.7 ± 18.6 | 45.3 ± 10.2 | 62.4 ± 18.6 | 46.2 ± 11.1 |
| 225 | 69.8 ± 23.1 | 42.5 ± 11.9 | 70.2 ± 22.8 | 44.0 ± 11.8 |
| 250 | 75.6 ± 23.2 | 43.9 ± 14.5 | 75.1 ± 23.8 | 44.3 ± 14.7 |
| 275 | 79.5 ± 22.2 | 50.6 ± 19.3 | 78.8 ± 22.9 | 50.2 ± 19.3 |
| 300 | 86.1 ± 20.5 | 57.0 ± 26.8 | 85.9 ± 20.5 | 56.5 ± 26.2 |
| 325 | 90.3 ± 16.9 | 66.2 ± 22.3 | 89.3 ± 18.2 | 65.5 ± 22.1 |
| 350 | 94.0 ± 13.8 | 77.1 ± 24.7 | 93.8 ± 13.9 | 76.7 ± 23.9 |
| 375 | 96.8 ± 8.9 | 82.9 ± 24.3 | 96.5 ± 9.3 | 82.7 ± 23.5 |
| 400 | 98.3 ± 4.6 | 90.4 ± 13.1 | 97.8 ± 4.6 | 89.1 ± 12.5 |