r1_c6
done · peak unseen_mismatch 49.9
α schedule
α 2.463 ×25 α 2.0338 ×25 α 1.7693 ×25 α 1.5668 ×25 α 1.3989 ×25 α 1.2536 ×25 α 1.1245 ×25 α 1.0077 ×25 α 0.9004 ×25 α 0.8011 ×25 α 0.7082 ×25 α 0.6209 ×25 α 0.5383 ×25 α 0.4599 ×25 α 0.3851 ×25 α 0.3136 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.3136
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 ± 1.4 | 50.3 ± 1.6 | 50.0 ± 1.5 | 49.3 ± 1.8 |
| 50 | 51.0 ± 1.3 | 50.3 ± 1.5 | 49.8 ± 1.3 | 49.9 ± 1.7 |
| 75 | 51.2 ± 1.7 | 49.7 ± 1.7 | 50.5 ± 1.5 | 49.4 ± 1.3 |
| 100 | 51.1 ± 2.7 | 50.2 ± 1.7 | 50.6 ± 1.7 | 49.7 ± 1.5 |
| 125 | 55.3 ± 8.2 | 48.7 ± 2.8 | 53.6 ± 8.7 | 48.7 ± 2.7 |
| 150 | 62.0 ± 16.2 | 49.4 ± 5.2 | 60.4 ± 16.1 | 48.2 ± 4.4 |
| 175 | 73.9 ± 21.4 | 47.3 ± 13.7 | 73.2 ± 21.8 | 46.9 ± 13.5 |
| 200 | 80.1 ± 22.3 | 47.7 ± 20.7 | 80.0 ± 22.3 | 47.5 ± 20.2 |
| 225 | 83.7 ± 22.1 | 44.1 ± 25.4 | 83.8 ± 22.0 | 45.0 ± 25.2 |
| 250 | 87.6 ± 18.9 | 40.6 ± 25.6 | 87.0 ± 19.7 | 40.8 ± 25.4 |
| 275 | 94.7 ± 12.4 | 42.5 ± 32.7 | 94.5 ± 12.6 | 42.7 ± 32.9 |
| 300 | 99.6 ± 1.6 | 40.6 ± 29.6 | 99.6 ± 1.7 | 40.3 ± 29.4 |
| 325 | 100.0 ± 0.0 | 33.0 ± 27.2 | 100.0 ± 0.0 | 33.1 ± 27.0 |
| 350 | 100.0 ± 0.0 | 29.9 ± 27.4 | 100.0 ± 0.0 | 30.0 ± 27.3 |
| 375 | 100.0 ± 0.0 | 27.7 ± 26.4 | 100.0 ± 0.0 | 28.2 ± 26.6 |
| 400 | 100.0 ± 0.0 | 27.9 ± 25.9 | 100.0 ± 0.0 | 28.2 ± 25.8 |