free-bdbc8cdc
done · peak unseen_mismatch 95.8
α schedule
α 2.73 ×26 α 1.132 ×11 α 2.55 ×32 α 1.538 ×28 α 2.842 ×13 α 1.431 ×21 α 3.32 ×41 α 1.389 ×19 α 3.298 ×33 α 0.9278 ×26 α 2.636 ×19 α 1.212 ×19 α 2.811 ×12 α 2.811 ×100
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 2.811
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 | 53.2 ± 4.5 | 49.6 ± 2.0 | 52.1 ± 4.3 | 49.6 ± 1.7 |
| 110 | 55.3 ± 7.8 | 50.1 ± 2.9 | 53.7 ± 7.8 | 49.6 ± 2.7 |
| 120 | 58.3 ± 10.7 | 49.8 ± 3.4 | 56.1 ± 10.5 | 49.8 ± 3.3 |
| 130 | 61.7 ± 14.4 | 51.4 ± 5.3 | 59.8 ± 14.1 | 50.6 ± 4.4 |
| 140 | 66.7 ± 17.2 | 54.0 ± 9.8 | 65.0 ± 17.3 | 52.2 ± 8.8 |
| 150 | 71.8 ± 19.0 | 58.1 ± 13.2 | 69.8 ± 19.4 | 55.8 ± 12.4 |
| 160 | 75.8 ± 19.1 | 62.8 ± 15.0 | 73.6 ± 19.7 | 60.2 ± 14.2 |
| 170 | 79.7 ± 19.2 | 68.8 ± 15.5 | 77.3 ± 19.9 | 66.0 ± 15.2 |
| 180 | 81.4 ± 19.3 | 70.4 ± 16.1 | 79.0 ± 20.0 | 67.3 ± 16.1 |
| 190 | 83.2 ± 19.3 | 69.5 ± 16.4 | 81.3 ± 20.3 | 67.2 ± 16.0 |
| 200 | 85.8 ± 18.1 | 73.6 ± 17.3 | 83.7 ± 19.8 | 70.3 ± 17.5 |
| 210 | 85.4 ± 17.8 | 75.5 ± 17.0 | 82.9 ± 19.5 | 72.2 ± 17.4 |
| 220 | 85.5 ± 17.1 | 74.9 ± 16.3 | 82.6 ± 17.9 | 70.6 ± 15.9 |
| 230 | 86.7 ± 16.0 | 77.3 ± 15.8 | 84.1 ± 17.4 | 73.3 ± 15.7 |
| 240 | 89.2 ± 15.1 | 76.3 ± 16.4 | 87.4 ± 16.1 | 72.9 ± 15.9 |
| 250 | 90.2 ± 15.4 | 75.3 ± 15.4 | 89.4 ± 15.6 | 72.9 ± 14.1 |
| 260 | 93.6 ± 11.7 | 81.6 ± 15.3 | 92.8 ± 12.6 | 79.3 ± 14.8 |
| 270 | 95.3 ± 8.3 | 80.1 ± 15.6 | 93.3 ± 10.4 | 77.5 ± 14.9 |
| 280 | 95.8 ± 6.9 | 78.3 ± 16.3 | 95.2 ± 7.4 | 76.3 ± 15.5 |
| 290 | 96.7 ± 6.7 | 73.0 ± 17.5 | 96.1 ± 7.6 | 72.7 ± 17.1 |
| 300 | 99.3 ± 1.1 | 73.7 ± 13.9 | 99.2 ± 1.4 | 74.4 ± 12.7 |
| 310 | 99.5 ± 0.8 | 75.8 ± 16.2 | 99.1 ± 1.2 | 74.2 ± 15.1 |
| 320 | 99.7 ± 0.5 | 81.2 ± 25.1 | 99.5 ± 1.0 | 80.5 ± 23.4 |
| 330 | 99.3 ± 0.1 | 65.8 ± 32.5 | 98.7 ± 0.4 | 66.4 ± 29.9 |
| 340 | 99.8 ± 0.0 | 49.8 ± 0.0 | 99.6 ± 0.0 | 50.0 ± 0.0 |
| 350 | 99.9 ± 0.0 | 61.9 ± 0.0 | 99.8 ± 0.0 | 60.8 ± 0.0 |
| 360 | 100.0 ± 0.0 | 72.3 ± 0.0 | 99.3 ± 0.0 | 70.8 ± 0.0 |
| 370 | 99.9 ± 0.0 | 84.8 ± 0.0 | 99.1 ± 0.0 | 80.6 ± 0.0 |
| 380 | 100.0 ± 0.0 | 80.3 ± 0.0 | 100.0 ± 0.0 | 76.4 ± 0.0 |
| 390 | 100.0 ± 0.0 | 90.9 ± 0.0 | 99.4 ± 0.0 | 88.6 ± 0.0 |
| 400 | 100.0 ± 0.0 | 97.1 ± 0.0 | 99.5 ± 0.0 | 95.8 ± 0.0 |