free-d59a33e7
done · peak unseen_mismatch 79.5
α schedule
α 4.238 ×54 α 0.8376 ×50 α 3.522 ×63 α 0.4329 ×33 α 3.305 ×30 α 1.129 ×16 α 3.423 ×54 α 3.423 ×100
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 3.423
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 | 53.9 ± 7.9 | 49.5 ± 2.6 | 52.7 ± 7.9 | 49.3 ± 2.3 |
| 110 | 55.7 ± 10.7 | 48.7 ± 4.7 | 54.7 ± 10.9 | 48.4 ± 4.9 |
| 120 | 58.8 ± 13.9 | 48.7 ± 5.9 | 57.9 ± 14.1 | 47.9 ± 5.9 |
| 130 | 62.2 ± 15.4 | 50.8 ± 7.7 | 60.3 ± 15.8 | 49.6 ± 7.2 |
| 140 | 67.0 ± 17.1 | 53.6 ± 10.7 | 64.8 ± 17.9 | 52.0 ± 10.0 |
| 150 | 71.1 ± 18.3 | 58.7 ± 12.7 | 68.4 ± 19.0 | 56.2 ± 12.0 |
| 160 | 74.3 ± 17.8 | 63.3 ± 12.5 | 71.3 ± 18.8 | 60.1 ± 11.4 |
| 170 | 78.9 ± 17.0 | 69.3 ± 14.4 | 75.7 ± 18.4 | 65.7 ± 14.3 |
| 180 | 82.5 ± 17.4 | 70.8 ± 16.5 | 80.4 ± 18.5 | 68.5 ± 16.1 |
| 190 | 86.7 ± 17.5 | 67.0 ± 17.2 | 85.7 ± 18.5 | 65.6 ± 16.8 |
| 200 | 89.6 ± 16.6 | 60.0 ± 18.7 | 88.7 ± 17.7 | 59.3 ± 18.3 |
| 210 | 92.2 ± 14.7 | 63.2 ± 18.5 | 91.3 ± 15.9 | 62.1 ± 18.7 |
| 220 | 94.7 ± 12.5 | 69.8 ± 20.1 | 93.7 ± 14.2 | 68.3 ± 20.0 |
| 230 | 95.5 ± 10.4 | 74.2 ± 18.5 | 94.4 ± 12.2 | 72.7 ± 18.3 |
| 240 | 95.8 ± 10.6 | 78.8 ± 19.1 | 94.6 ± 12.5 | 77.2 ± 19.2 |
| 250 | 96.5 ± 10.4 | 78.1 ± 19.2 | 95.3 ± 12.2 | 76.5 ± 19.0 |
| 260 | 95.5 ± 11.6 | 74.6 ± 19.2 | 94.0 ± 13.4 | 72.7 ± 18.9 |
| 270 | 92.6 ± 13.7 | 71.3 ± 15.5 | 90.4 ± 16.2 | 69.3 ± 15.1 |
| 280 | 93.0 ± 14.2 | 77.5 ± 16.6 | 91.6 ± 15.7 | 75.1 ± 15.4 |
| 290 | 91.7 ± 15.6 | 76.6 ± 17.3 | 89.5 ± 15.8 | 74.2 ± 16.0 |
| 300 | 90.6 ± 15.1 | 80.2 ± 19.1 | 88.2 ± 16.2 | 76.9 ± 18.0 |
| 310 | 88.9 ± 11.5 | 79.0 ± 9.4 | 83.5 ± 14.9 | 72.5 ± 9.4 |
| 320 | 89.5 ± 9.5 | 85.8 ± 8.6 | 83.8 ± 13.1 | 79.5 ± 9.8 |
| 330 | 87.4 ± 5.1 | 84.9 ± 4.3 | 76.8 ± 8.5 | 75.1 ± 5.4 |
| 340 | 89.3 ± 4.6 | 86.3 ± 3.6 | 79.2 ± 7.0 | 78.0 ± 3.4 |
| 350 | 88.9 ± 4.4 | 87.3 ± 1.6 | 79.5 ± 4.4 | 77.9 ± 2.2 |
| 360 | 88.6 ± 2.6 | 87.3 ± 1.3 | 78.7 ± 2.8 | 76.5 ± 1.5 |
| 370 | 89.7 ± 2.4 | 87.6 ± 1.1 | 80.8 ± 2.5 | 78.3 ± 1.2 |
| 380 | 87.9 ± 2.0 | 86.5 ± 1.6 | 77.9 ± 0.8 | 76.0 ± 1.9 |
| 390 | 89.6 ± 1.5 | 87.6 ± 1.3 | 80.1 ± 0.9 | 77.8 ± 3.8 |
| 400 | 89.0 ± 0.2 | 87.7 ± 2.1 | 76.8 ± 1.8 | 75.1 ± 4.2 |