free-36f2d865
done · peak unseen_mismatch 86.1
α schedule
α 4.193 ×30 α 1.356 ×22 α 4.456 ×36 α 1.8 ×21 α 4.31 ×19 α 0.9352 ×25 α 4.117 ×19 α 1.41 ×50 α 4.477 ×12 α 1.297 ×21 α 4.84 ×13 α 1.68 ×20 α 4.56 ×12 α 4.56 ×100
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 4.56
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 | 55.5 ± 5.9 | 51.1 ± 2.8 | 52.7 ± 5.4 | 49.6 ± 2.2 |
| 110 | 58.6 ± 10.1 | 52.4 ± 4.9 | 55.6 ± 9.3 | 50.8 ± 3.6 |
| 120 | 62.8 ± 12.9 | 55.9 ± 7.0 | 59.3 ± 12.5 | 52.7 ± 6.0 |
| 130 | 67.5 ± 14.9 | 59.8 ± 9.7 | 63.2 ± 14.8 | 56.2 ± 8.7 |
| 140 | 71.3 ± 17.3 | 62.4 ± 12.6 | 67.8 ± 17.6 | 58.6 ± 12.1 |
| 150 | 75.7 ± 18.8 | 64.2 ± 14.6 | 73.7 ± 19.2 | 61.4 ± 14.1 |
| 160 | 81.6 ± 18.6 | 68.6 ± 17.2 | 79.8 ± 19.2 | 65.8 ± 16.9 |
| 170 | 83.4 ± 17.8 | 73.2 ± 17.0 | 81.2 ± 18.4 | 70.0 ± 16.2 |
| 180 | 85.6 ± 17.7 | 75.3 ± 18.3 | 83.4 ± 18.2 | 72.1 ± 17.4 |
| 190 | 87.4 ± 17.5 | 73.4 ± 18.5 | 85.7 ± 18.5 | 70.8 ± 17.9 |
| 200 | 90.0 ± 15.6 | 74.6 ± 16.4 | 88.3 ± 17.3 | 72.5 ± 16.2 |
| 210 | 92.6 ± 12.9 | 74.7 ± 16.0 | 91.1 ± 14.8 | 72.5 ± 15.4 |
| 220 | 94.5 ± 11.4 | 78.0 ± 15.7 | 93.5 ± 12.3 | 76.2 ± 15.2 |
| 230 | 95.4 ± 10.8 | 80.1 ± 15.6 | 94.6 ± 11.0 | 78.8 ± 14.9 |
| 240 | 95.1 ± 11.6 | 74.5 ± 19.7 | 94.1 ± 12.2 | 72.6 ± 19.3 |
| 250 | 95.6 ± 11.5 | 73.4 ± 24.2 | 94.5 ± 12.6 | 71.8 ± 24.4 |
| 260 | 95.2 ± 11.7 | 66.8 ± 23.0 | 94.2 ± 12.6 | 65.1 ± 22.4 |
| 270 | 94.8 ± 12.3 | 72.1 ± 21.3 | 94.0 ± 12.7 | 69.8 ± 20.6 |
| 280 | 96.4 ± 8.3 | 80.5 ± 16.0 | 95.8 ± 9.3 | 77.0 ± 15.8 |
| 290 | 97.1 ± 5.5 | 77.5 ± 18.2 | 96.3 ± 6.9 | 75.5 ± 17.8 |
| 300 | 97.5 ± 4.7 | 84.2 ± 19.8 | 96.2 ± 6.2 | 82.0 ± 19.7 |
| 310 | 93.8 ± 6.0 | 67.2 ± 17.0 | 91.7 ± 8.3 | 62.3 ± 13.4 |
| 320 | 96.0 ± 4.0 | 70.5 ± 18.4 | 92.8 ± 7.2 | 65.4 ± 14.7 |
| 330 | 96.0 ± 4.0 | 85.2 ± 4.0 | 92.5 ± 7.5 | 79.8 ± 0.1 |
| 340 | 95.7 ± 4.0 | 77.5 ± 13.2 | 94.2 ± 5.7 | 74.4 ± 11.8 |
| 350 | 95.8 ± 4.2 | 80.6 ± 8.0 | 92.5 ± 7.5 | 75.6 ± 6.8 |
| 360 | 95.9 ± 3.9 | 87.0 ± 3.4 | 93.0 ± 6.8 | 78.7 ± 3.4 |
| 370 | 94.8 ± 5.0 | 87.8 ± 1.2 | 91.7 ± 7.4 | 81.0 ± 2.4 |
| 380 | 96.0 ± 2.8 | 91.9 ± 3.0 | 92.0 ± 5.5 | 86.1 ± 6.4 |
| 390 | 94.2 ± 0.0 | 90.5 ± 0.0 | 87.6 ± 0.0 | 81.8 ± 0.0 |
| 400 | 95.1 ± 0.0 | 92.1 ± 0.0 | 89.3 ± 0.0 | 84.8 ± 0.0 |