free-9160bea0
done · peak unseen_mismatch 93.0
α schedule
α 4.371 ×38 α 1.479 ×23 α 3.626 ×45 α 1.524 ×18 α 3.86 ×59 α 0.9806 ×61 α 3.74 ×19 α 1.129 ×26 α 4.04 ×11 α 4.04 ×100
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 4.04
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 | 56.4 ± 7.8 | 52.0 ± 3.5 | 53.0 ± 7.0 | 50.9 ± 3.1 |
| 110 | 59.7 ± 10.1 | 55.0 ± 5.9 | 55.7 ± 9.1 | 51.8 ± 4.5 |
| 120 | 63.3 ± 12.6 | 57.4 ± 8.3 | 58.5 ± 12.7 | 54.0 ± 6.9 |
| 130 | 69.3 ± 14.9 | 62.5 ± 10.9 | 64.3 ± 15.5 | 57.8 ± 10.3 |
| 140 | 75.5 ± 15.3 | 69.5 ± 13.3 | 69.6 ± 16.9 | 63.1 ± 13.4 |
| 150 | 79.5 ± 14.2 | 74.1 ± 13.3 | 72.8 ± 15.3 | 66.7 ± 13.1 |
| 160 | 83.0 ± 12.2 | 77.8 ± 11.9 | 75.5 ± 13.1 | 69.6 ± 11.6 |
| 170 | 85.3 ± 10.6 | 81.5 ± 10.6 | 77.0 ± 12.1 | 72.3 ± 10.8 |
| 180 | 86.3 ± 8.6 | 83.2 ± 8.7 | 76.8 ± 10.8 | 72.7 ± 9.5 |
| 190 | 89.0 ± 8.3 | 86.2 ± 8.4 | 81.1 ± 10.9 | 77.2 ± 9.9 |
| 200 | 94.2 ± 8.8 | 91.6 ± 9.9 | 90.3 ± 11.3 | 86.8 ± 11.4 |
| 210 | 94.8 ± 9.6 | 90.7 ± 11.7 | 91.6 ± 11.5 | 86.7 ± 11.8 |
| 220 | 94.5 ± 9.9 | 87.1 ± 14.7 | 91.8 ± 11.8 | 84.1 ± 14.1 |
| 230 | 94.1 ± 10.3 | 78.9 ± 11.7 | 92.9 ± 11.3 | 77.1 ± 11.1 |
| 240 | 95.2 ± 9.4 | 81.2 ± 11.5 | 93.8 ± 11.0 | 79.3 ± 11.8 |
| 250 | 97.9 ± 4.0 | 75.4 ± 14.0 | 96.4 ± 6.3 | 73.6 ± 12.6 |
| 260 | 97.9 ± 3.6 | 80.5 ± 12.6 | 96.4 ± 6.1 | 79.6 ± 11.6 |
| 270 | 98.1 ± 2.6 | 88.1 ± 3.9 | 97.3 ± 3.8 | 85.1 ± 2.4 |
| 280 | 98.6 ± 1.9 | 94.7 ± 1.3 | 99.3 ± 1.0 | 93.0 ± 1.3 |