fixed_1.6
done · peak unseen_mismatch 62.9
α schedule
α 1.6 ×400
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.6
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 | 51.4 ± 2.2 | 49.8 ± 1.0 | 50.3 ± 1.7 | 49.9 ± 1.9 |
| 110 | 52.1 ± 2.0 | 49.4 ± 1.5 | 50.6 ± 1.7 | 49.8 ± 1.7 |
| 120 | 51.9 ± 2.9 | 49.7 ± 2.1 | 51.6 ± 2.3 | 49.2 ± 1.3 |
| 130 | 52.0 ± 3.0 | 50.0 ± 1.6 | 51.7 ± 2.5 | 49.5 ± 1.8 |
| 140 | 52.6 ± 3.6 | 49.4 ± 2.5 | 52.9 ± 4.2 | 49.1 ± 1.7 |
| 150 | 54.3 ± 6.9 | 48.5 ± 3.8 | 54.7 ± 6.9 | 48.0 ± 4.0 |
| 160 | 57.5 ± 11.3 | 48.9 ± 4.4 | 56.8 ± 11.4 | 48.6 ± 3.8 |
| 170 | 61.1 ± 15.1 | 50.3 ± 7.8 | 60.3 ± 15.0 | 49.9 ± 7.1 |
| 180 | 64.9 ± 18.0 | 49.8 ± 13.8 | 64.2 ± 18.0 | 48.9 ± 13.3 |
| 190 | 71.7 ± 19.5 | 51.3 ± 12.6 | 70.0 ± 20.2 | 49.8 ± 12.3 |
| 200 | 77.1 ± 20.4 | 51.6 ± 13.4 | 75.8 ± 21.1 | 50.5 ± 13.8 |
| 210 | 80.7 ± 20.9 | 52.1 ± 15.5 | 80.0 ± 21.2 | 50.5 ± 15.5 |
| 220 | 85.4 ± 20.3 | 50.4 ± 18.1 | 84.5 ± 21.0 | 49.2 ± 17.9 |
| 230 | 87.2 ± 19.5 | 50.6 ± 22.7 | 86.8 ± 19.8 | 49.5 ± 22.2 |
| 240 | 89.4 ± 17.1 | 51.3 ± 22.1 | 89.0 ± 17.5 | 50.3 ± 21.5 |
| 250 | 92.1 ± 15.9 | 55.2 ± 19.3 | 92.1 ± 15.7 | 54.0 ± 19.4 |
| 260 | 93.4 ± 15.2 | 59.0 ± 18.2 | 93.3 ± 15.6 | 57.8 ± 18.5 |
| 270 | 93.5 ± 15.3 | 54.8 ± 25.7 | 93.6 ± 15.0 | 54.5 ± 25.2 |
| 280 | 94.1 ± 14.6 | 54.1 ± 19.6 | 93.7 ± 15.5 | 52.5 ± 19.9 |
| 290 | 94.6 ± 13.8 | 57.6 ± 20.2 | 93.9 ± 15.4 | 55.5 ± 20.6 |
| 300 | 95.3 ± 11.8 | 57.0 ± 17.1 | 94.2 ± 13.6 | 55.0 ± 17.2 |
| 310 | 95.7 ± 11.7 | 57.4 ± 23.8 | 94.7 ± 13.0 | 56.5 ± 23.1 |
| 320 | 96.6 ± 10.0 | 54.0 ± 28.7 | 96.3 ± 10.4 | 53.0 ± 28.4 |
| 330 | 98.1 ± 5.5 | 61.1 ± 25.6 | 97.7 ± 5.9 | 59.0 ± 24.5 |
| 340 | 98.9 ± 3.0 | 64.1 ± 28.6 | 98.5 ± 3.4 | 62.9 ± 28.5 |
| 350 | 99.4 ± 1.1 | 60.0 ± 21.9 | 99.0 ± 1.8 | 58.6 ± 21.4 |
| 360 | 99.9 ± 0.1 | 48.7 ± 16.6 | 100.0 ± 0.0 | 48.0 ± 17.3 |
| 370 | 100.0 ± 0.0 | 50.4 ± 13.5 | 100.0 ± 0.0 | 50.0 ± 14.7 |
| 380 | 100.0 ± 0.0 | 51.1 ± 20.2 | 100.0 ± 0.0 | 51.8 ± 20.3 |
| 390 | 100.0 ± 0.0 | 55.4 ± 19.5 | 100.0 ± 0.0 | 55.1 ± 19.0 |
| 400 | 100.0 ± 0.0 | 61.2 ± 26.2 | 100.0 ± 0.0 | 59.4 ± 26.3 |