fixed_1.5
done · peak unseen_mismatch 60.6
α schedule
α 1.5 ×400
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.5
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 |
|---|---|---|---|---|
| 100 | 50.9 ± 1.6 | 50.1 ± 1.4 | 50.0 ± 2.0 | 50.0 ± 2.0 |
| 110 | 51.4 ± 1.6 | 49.9 ± 1.7 | 50.5 ± 1.5 | 50.0 ± 1.6 |
| 120 | 51.4 ± 2.0 | 49.6 ± 1.7 | 51.3 ± 2.5 | 49.0 ± 1.5 |
| 130 | 51.4 ± 2.6 | 49.7 ± 1.9 | 51.4 ± 2.3 | 49.4 ± 1.7 |
| 140 | 52.1 ± 3.1 | 49.1 ± 2.1 | 52.2 ± 2.9 | 48.9 ± 2.1 |
| 150 | 54.1 ± 6.8 | 48.6 ± 3.1 | 54.2 ± 6.9 | 48.3 ± 2.6 |
| 160 | 56.1 ± 10.6 | 47.6 ± 5.2 | 55.4 ± 11.0 | 47.5 ± 5.1 |
| 170 | 58.4 ± 13.6 | 47.3 ± 7.8 | 57.3 ± 13.9 | 47.0 ± 7.5 |
| 180 | 63.1 ± 16.0 | 45.9 ± 11.8 | 62.6 ± 16.0 | 45.5 ± 11.6 |
| 190 | 68.3 ± 20.3 | 48.3 ± 8.0 | 67.3 ± 20.3 | 47.0 ± 7.7 |
| 200 | 71.6 ± 21.9 | 47.4 ± 13.9 | 70.5 ± 22.6 | 47.0 ± 14.5 |
| 210 | 74.6 ± 21.9 | 46.9 ± 13.6 | 73.6 ± 22.6 | 46.4 ± 14.0 |
| 220 | 79.4 ± 20.0 | 45.6 ± 15.0 | 78.4 ± 20.8 | 44.9 ± 15.2 |
| 230 | 82.2 ± 20.2 | 46.7 ± 16.9 | 81.7 ± 20.5 | 46.3 ± 16.4 |
| 240 | 87.7 ± 17.1 | 44.6 ± 16.9 | 87.1 ± 17.4 | 44.5 ± 17.1 |
| 250 | 90.6 ± 15.4 | 42.8 ± 20.5 | 90.3 ± 15.7 | 42.7 ± 20.8 |
| 260 | 94.3 ± 11.8 | 40.0 ± 24.2 | 93.9 ± 12.5 | 39.0 ± 24.2 |
| 270 | 96.7 ± 9.0 | 39.5 ± 23.9 | 96.6 ± 8.9 | 39.1 ± 23.3 |
| 280 | 98.7 ± 4.1 | 46.1 ± 21.0 | 98.7 ± 4.0 | 45.7 ± 20.8 |
| 290 | 99.8 ± 0.7 | 54.0 ± 23.5 | 99.8 ± 0.5 | 53.1 ± 23.4 |
| 300 | 100.0 ± 0.0 | 53.0 ± 26.7 | 100.0 ± 0.1 | 52.2 ± 26.2 |
| 310 | 100.0 ± 0.0 | 56.8 ± 24.4 | 100.0 ± 0.1 | 55.6 ± 24.2 |
| 320 | 100.0 ± 0.0 | 55.2 ± 22.4 | 100.0 ± 0.0 | 54.7 ± 22.1 |
| 330 | 100.0 ± 0.0 | 53.0 ± 25.3 | 100.0 ± 0.1 | 52.2 ± 25.5 |
| 340 | 100.0 ± 0.0 | 52.4 ± 22.9 | 100.0 ± 0.0 | 52.1 ± 22.6 |
| 350 | 100.0 ± 0.0 | 55.5 ± 26.0 | 100.0 ± 0.0 | 55.2 ± 25.5 |
| 360 | 100.0 ± 0.1 | 60.5 ± 25.5 | 99.9 ± 0.3 | 60.6 ± 26.1 |
| 370 | 100.0 ± 0.1 | 58.8 ± 23.8 | 99.9 ± 0.2 | 59.0 ± 23.7 |
| 380 | 100.0 ± 0.0 | 52.8 ± 28.1 | 100.0 ± 0.0 | 53.0 ± 27.6 |
| 390 | 100.0 ± 0.0 | 46.1 ± 19.8 | 100.0 ± 0.0 | 46.7 ± 19.7 |
| 400 | 100.0 ± 0.0 | 49.1 ± 21.8 | 100.0 ± 0.0 | 49.8 ± 21.0 |