free-152ceb4f
done · peak unseen_mismatch 82.8
α schedule
α 2.925 ×35 α 0.7198 ×13 α 3.771 ×91 α 0.03986 ×14 α 4.005 ×14 α 0.0941 ×17 α 3.373 ×67 α 0.3445 ×49 α 0.3445 ×100
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.3445
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 | 53.0 ± 3.7 | 50.6 ± 2.3 | 51.3 ± 2.8 | 49.5 ± 1.6 |
| 110 | 55.5 ± 6.4 | 52.2 ± 4.5 | 52.6 ± 5.7 | 50.6 ± 3.3 |
| 120 | 59.3 ± 9.2 | 54.7 ± 7.0 | 55.1 ± 8.5 | 51.6 ± 4.9 |
| 130 | 63.8 ± 12.7 | 59.1 ± 9.8 | 58.7 ± 11.8 | 54.5 ± 8.1 |
| 140 | 68.2 ± 14.0 | 62.8 ± 11.6 | 62.2 ± 13.3 | 57.2 ± 9.9 |
| 150 | 69.5 ± 16.6 | 63.0 ± 13.6 | 65.9 ± 16.5 | 59.6 ± 12.3 |
| 160 | 75.6 ± 18.5 | 66.9 ± 16.6 | 72.9 ± 19.6 | 63.9 ± 15.9 |
| 170 | 77.4 ± 17.8 | 68.5 ± 16.3 | 74.0 ± 18.9 | 64.9 ± 15.7 |
| 180 | 80.8 ± 18.5 | 69.2 ± 17.5 | 79.0 ± 19.0 | 66.8 ± 16.7 |
| 190 | 85.5 ± 17.0 | 69.4 ± 16.4 | 83.8 ± 18.4 | 66.6 ± 16.1 |
| 200 | 89.3 ± 14.5 | 74.2 ± 16.7 | 87.8 ± 15.6 | 71.4 ± 16.6 |
| 210 | 91.2 ± 13.4 | 78.3 ± 16.3 | 89.1 ± 15.2 | 74.7 ± 15.9 |
| 220 | 91.9 ± 12.4 | 80.3 ± 14.7 | 88.9 ± 14.4 | 76.0 ± 14.5 |
| 230 | 91.3 ± 12.3 | 80.0 ± 14.4 | 87.4 ± 15.1 | 74.8 ± 15.0 |
| 240 | 91.4 ± 10.1 | 82.3 ± 12.6 | 86.5 ± 13.5 | 75.9 ± 13.1 |
| 250 | 92.7 ± 7.3 | 88.3 ± 7.6 | 87.0 ± 12.0 | 80.9 ± 10.4 |
| 260 | 93.7 ± 7.4 | 88.5 ± 9.2 | 89.4 ± 10.3 | 82.8 ± 10.6 |
| 270 | 95.7 ± 6.3 | 83.9 ± 17.5 | 93.6 ± 8.2 | 81.1 ± 16.9 |
| 280 | 97.8 ± 2.5 | 78.3 ± 20.6 | 97.6 ± 3.3 | 75.7 ± 20.3 |
| 290 | 99.0 ± 1.6 | 67.0 ± 23.8 | 98.8 ± 2.0 | 65.4 ± 23.8 |
| 300 | 100.0 ± 0.0 | 60.6 ± 19.8 | 100.0 ± 0.0 | 59.5 ± 19.1 |
| 310 | 100.0 ± 0.0 | 51.5 ± 35.0 | 100.0 ± 0.0 | 51.9 ± 33.8 |
| 320 | 100.0 ± 0.0 | 25.9 ± 13.8 | 100.0 ± 0.0 | 27.8 ± 13.8 |
| 330 | 100.0 ± 0.0 | 28.2 ± 20.9 | 100.0 ± 0.0 | 29.0 ± 22.0 |
| 340 | 100.0 ± 0.0 | 25.4 ± 23.1 | 100.0 ± 0.0 | 25.6 ± 23.8 |
| 350 | 100.0 ± 0.0 | 21.1 ± 18.4 | 100.0 ± 0.0 | 21.4 ± 19.1 |
| 360 | 100.0 ± 0.0 | 3.5 ± 2.9 | 100.0 ± 0.0 | 3.0 ± 2.9 |
| 370 | 100.0 ± 0.0 | 12.2 ± 11.8 | 100.0 ± 0.0 | 11.9 ± 11.2 |
| 380 | 100.0 ± 0.0 | 14.8 ± 14.1 | 100.0 ± 0.0 | 15.3 ± 14.3 |
| 390 | 100.0 ± 0.0 | 18.7 ± 16.6 | 100.0 ± 0.0 | 19.0 ± 17.4 |
| 400 | 100.0 ± 0.0 | 18.6 ± 1.6 | 100.0 ± 0.0 | 18.9 ± 3.0 |