free-173cb72a
done · peak unseen_mismatch 93.7
α schedule
α 4.636 ×21 α 1.407 ×18 α 4.531 ×27 α 1.203 ×34 α 4.833 ×19 α 1.189 ×15 α 5.0 ×35 α 1.085 ×13 α 4.334 ×19 α 0.7515 ×14 α 5.0 ×59 α 1.362 ×26 α 1.362 ×100
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.362
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 | 52.9 ± 6.1 | 49.5 ± 1.9 | 51.6 ± 6.2 | 50.3 ± 1.6 |
| 110 | 54.8 ± 7.4 | 50.3 ± 3.4 | 53.4 ± 7.3 | 50.0 ± 3.2 |
| 120 | 57.5 ± 10.1 | 51.5 ± 4.8 | 55.5 ± 10.1 | 50.5 ± 4.5 |
| 130 | 61.6 ± 13.3 | 53.2 ± 7.6 | 59.6 ± 13.4 | 51.4 ± 6.9 |
| 140 | 67.4 ± 16.4 | 55.8 ± 9.8 | 65.2 ± 16.8 | 53.8 ± 9.3 |
| 150 | 73.4 ± 17.4 | 61.2 ± 11.6 | 70.5 ± 18.4 | 57.9 ± 11.3 |
| 160 | 78.3 ± 17.2 | 67.4 ± 13.2 | 75.0 ± 18.7 | 63.5 ± 13.4 |
| 170 | 81.5 ± 16.5 | 74.0 ± 14.2 | 77.9 ± 18.1 | 69.7 ± 14.7 |
| 180 | 84.5 ± 16.2 | 75.5 ± 14.0 | 81.7 ± 17.8 | 71.6 ± 14.8 |
| 190 | 87.9 ± 14.0 | 79.4 ± 13.3 | 85.1 ± 15.7 | 75.3 ± 14.2 |
| 200 | 90.3 ± 11.0 | 83.3 ± 12.0 | 86.3 ± 13.7 | 78.0 ± 13.5 |
| 210 | 92.0 ± 10.5 | 83.9 ± 13.1 | 88.8 ± 12.5 | 79.3 ± 13.5 |
| 220 | 94.3 ± 11.0 | 83.8 ± 14.0 | 91.8 ± 12.8 | 81.1 ± 13.9 |
| 230 | 94.1 ± 10.4 | 85.4 ± 13.2 | 90.2 ± 13.5 | 80.8 ± 13.1 |
| 240 | 91.4 ± 11.9 | 85.5 ± 11.3 | 86.3 ± 13.5 | 79.3 ± 11.1 |
| 250 | 90.1 ± 12.1 | 85.0 ± 12.1 | 83.8 ± 13.2 | 77.7 ± 11.4 |
| 260 | 89.1 ± 11.5 | 84.2 ± 10.3 | 81.7 ± 12.0 | 75.2 ± 8.8 |
| 270 | 88.7 ± 11.4 | 85.7 ± 10.3 | 80.9 ± 11.7 | 77.0 ± 11.1 |
| 280 | 89.0 ± 13.2 | 87.7 ± 11.6 | 82.3 ± 13.0 | 79.6 ± 12.4 |
| 290 | 91.3 ± 15.1 | 90.0 ± 13.9 | 87.7 ± 13.8 | 84.8 ± 13.9 |
| 300 | 91.0 ± 15.7 | 88.2 ± 17.7 | 86.7 ± 16.6 | 83.7 ± 17.0 |
| 310 | 82.0 ± 22.4 | 79.3 ± 23.4 | 79.1 ± 21.5 | 76.2 ± 19.7 |
| 320 | 75.6 ± 21.4 | 70.4 ± 24.1 | 71.8 ± 19.6 | 69.2 ± 21.2 |
| 330 | 77.0 ± 22.1 | 73.0 ± 21.8 | 74.0 ± 19.9 | 71.4 ± 20.9 |
| 340 | 77.3 ± 21.4 | 71.6 ± 20.3 | 73.3 ± 19.4 | 70.0 ± 16.4 |
| 350 | 82.0 ± 17.1 | 78.4 ± 18.2 | 80.0 ± 16.1 | 74.2 ± 16.7 |
| 360 | 88.2 ± 11.4 | 81.2 ± 15.4 | 86.3 ± 11.1 | 78.1 ± 12.0 |
| 370 | 98.6 ± 1.0 | 95.5 ± 3.1 | 97.2 ± 0.4 | 93.7 ± 0.7 |