free-b261320b
done · peak unseen_mismatch 79.8
α schedule
α 4.245 ×61 α 1.613 ×94 α 4.384 ×25 α 1.112 ×53 α 4.671 ×67 α 4.671 ×100
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 4.671
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 | 55.7 ± 7.5 | 52.1 ± 4.6 | 52.9 ± 7.0 | 50.7 ± 3.5 |
| 110 | 58.4 ± 10.2 | 53.7 ± 7.2 | 54.9 ± 9.4 | 51.4 ± 5.5 |
| 120 | 62.3 ± 13.6 | 56.4 ± 10.1 | 58.9 ± 13.2 | 53.7 ± 8.8 |
| 130 | 67.8 ± 16.9 | 60.1 ± 13.0 | 64.7 ± 16.3 | 56.8 ± 11.8 |
| 140 | 72.2 ± 18.4 | 63.2 ± 15.1 | 69.6 ± 18.5 | 59.9 ± 13.8 |
| 150 | 78.1 ± 18.4 | 66.9 ± 16.3 | 75.7 ± 19.2 | 63.6 ± 15.6 |
| 160 | 82.2 ± 17.7 | 69.8 ± 16.4 | 79.9 ± 18.6 | 66.7 ± 15.9 |
| 170 | 86.3 ± 16.4 | 76.6 ± 16.4 | 83.5 ± 17.0 | 72.7 ± 15.8 |
| 180 | 87.6 ± 14.9 | 80.0 ± 14.6 | 84.2 ± 15.7 | 75.9 ± 14.7 |
| 190 | 88.9 ± 14.9 | 78.9 ± 15.0 | 86.3 ± 15.2 | 75.2 ± 14.0 |
| 200 | 91.8 ± 13.9 | 81.2 ± 14.6 | 90.5 ± 14.7 | 78.8 ± 14.2 |
| 210 | 93.5 ± 13.5 | 80.1 ± 16.8 | 92.4 ± 14.0 | 77.5 ± 16.6 |
| 220 | 93.9 ± 15.2 | 75.1 ± 18.0 | 93.4 ± 15.7 | 73.3 ± 16.7 |
| 230 | 94.2 ± 15.4 | 71.3 ± 22.0 | 94.0 ± 15.9 | 69.6 ± 21.1 |
| 240 | 93.4 ± 15.0 | 67.6 ± 21.4 | 93.0 ± 15.3 | 67.0 ± 20.5 |
| 250 | 93.6 ± 14.3 | 73.2 ± 19.7 | 92.5 ± 16.1 | 71.7 ± 19.1 |
| 260 | 90.8 ± 17.2 | 66.7 ± 15.8 | 90.0 ± 17.4 | 65.2 ± 14.9 |
| 270 | 91.3 ± 17.7 | 74.3 ± 15.6 | 91.0 ± 17.2 | 73.1 ± 15.1 |
| 280 | 91.0 ± 16.7 | 77.9 ± 18.2 | 90.0 ± 17.6 | 75.9 ± 17.2 |
| 290 | 90.5 ± 16.3 | 79.0 ± 19.0 | 87.5 ± 19.6 | 76.6 ± 20.1 |
| 300 | 85.6 ± 16.8 | 69.7 ± 15.7 | 79.4 ± 22.0 | 64.9 ± 16.0 |
| 310 | 86.6 ± 15.3 | 79.4 ± 11.9 | 79.0 ± 21.6 | 72.7 ± 14.4 |
| 320 | 85.7 ± 14.6 | 84.0 ± 14.4 | 79.8 ± 20.1 | 79.8 ± 19.3 |
| 330 | 79.9 ± 13.1 | 78.5 ± 11.7 | 71.5 ± 18.1 | 70.8 ± 17.2 |
| 340 | 81.2 ± 13.2 | 79.8 ± 10.9 | 72.8 ± 18.2 | 70.5 ± 17.4 |
| 350 | 82.9 ± 9.8 | 80.0 ± 7.7 | 72.1 ± 17.4 | 69.0 ± 15.7 |
| 360 | 83.0 ± 9.8 | 81.7 ± 7.6 | 73.0 ± 15.3 | 70.8 ± 13.3 |
| 370 | 83.8 ± 9.3 | 82.5 ± 7.2 | 71.5 ± 16.7 | 71.0 ± 14.4 |
| 380 | 82.8 ± 8.5 | 81.6 ± 6.4 | 71.0 ± 13.8 | 71.2 ± 12.3 |
| 390 | 83.5 ± 8.1 | 82.4 ± 6.2 | 71.8 ± 12.4 | 71.5 ± 11.6 |
| 400 | 84.5 ± 7.1 | 83.0 ± 5.3 | 73.0 ± 12.8 | 72.8 ± 10.2 |