free-3c84b515
done · peak unseen_mismatch 85.9
α schedule
α 3.593 ×52 α 4.019 ×28 α 0.9766 ×11 α 0.2553 ×13 α 3.218 ×19 α 1.084 ×30 α 4.947 ×46 α 3.688 ×11 α 1.443 ×17 α 3.728 ×12 α 2.472 ×15 α 0.7549 ×20 α 3.462 ×15 α 4.598 ×11 α 4.598 ×100
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 4.598
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 | 54.3 ± 5.1 | 51.1 ± 2.3 | 52.3 ± 4.4 | 49.8 ± 2.2 |
| 110 | 56.7 ± 8.6 | 51.3 ± 3.0 | 54.5 ± 7.9 | 50.0 ± 2.5 |
| 120 | 61.8 ± 13.3 | 54.6 ± 7.4 | 58.8 ± 13.1 | 51.8 ± 5.9 |
| 130 | 65.9 ± 15.4 | 57.0 ± 9.6 | 62.8 ± 15.4 | 54.4 ± 8.7 |
| 140 | 71.1 ± 18.3 | 60.1 ± 11.9 | 68.5 ± 19.0 | 57.5 ± 11.2 |
| 150 | 75.2 ± 19.2 | 58.9 ± 13.2 | 73.6 ± 19.9 | 56.5 ± 12.8 |
| 160 | 81.5 ± 18.6 | 64.7 ± 15.1 | 80.0 ± 19.4 | 61.7 ± 14.8 |
| 170 | 86.1 ± 16.8 | 72.6 ± 16.7 | 84.1 ± 17.8 | 69.7 ± 16.0 |
| 180 | 88.8 ± 15.4 | 78.3 ± 15.3 | 86.4 ± 16.4 | 74.8 ± 14.7 |
| 190 | 90.0 ± 14.0 | 81.1 ± 14.6 | 87.2 ± 15.1 | 77.1 ± 14.0 |
| 200 | 90.2 ± 12.2 | 84.0 ± 11.9 | 86.2 ± 14.5 | 78.5 ± 12.7 |
| 210 | 90.5 ± 11.4 | 85.5 ± 11.4 | 85.7 ± 13.9 | 79.5 ± 12.6 |
| 220 | 91.9 ± 11.7 | 88.2 ± 11.8 | 88.5 ± 13.7 | 82.9 ± 13.4 |
| 230 | 92.9 ± 10.6 | 87.0 ± 13.0 | 89.7 ± 13.6 | 81.9 ± 14.8 |
| 240 | 91.3 ± 9.5 | 87.8 ± 9.0 | 85.9 ± 12.5 | 80.4 ± 11.1 |
| 250 | 91.7 ± 8.6 | 87.7 ± 7.6 | 85.0 ± 12.0 | 79.9 ± 9.4 |
| 260 | 94.0 ± 8.8 | 91.3 ± 8.2 | 90.0 ± 11.9 | 85.6 ± 10.5 |
| 270 | 95.0 ± 11.2 | 90.4 ± 10.5 | 92.6 ± 14.0 | 85.9 ± 12.5 |
| 280 | 94.0 ± 12.5 | 83.8 ± 15.4 | 91.4 ± 17.7 | 79.8 ± 18.3 |
| 290 | 90.7 ± 13.2 | 82.3 ± 10.5 | 86.3 ± 19.3 | 76.5 ± 13.8 |
| 300 | 85.5 ± 13.9 | 82.8 ± 11.0 | 80.5 ± 19.1 | 73.8 ± 12.2 |
| 310 | 87.5 ± 11.1 | 75.8 ± 1.1 | 80.0 ± 16.9 | 67.0 ± 1.9 |
| 320 | 89.5 ± 9.8 | 86.5 ± 6.8 | 81.8 ± 16.1 | 74.3 ± 6.3 |
| 330 | 90.2 ± 8.1 | 84.6 ± 5.0 | 81.7 ± 13.4 | 73.4 ± 4.6 |
| 340 | 90.0 ± 7.0 | 86.9 ± 4.1 | 79.7 ± 13.8 | 74.3 ± 4.7 |
| 350 | 90.4 ± 6.8 | 89.3 ± 5.6 | 81.5 ± 13.2 | 76.2 ± 6.1 |
| 360 | 90.3 ± 6.0 | 89.5 ± 5.2 | 81.8 ± 11.9 | 77.8 ± 6.8 |
| 370 | 90.4 ± 5.4 | 89.3 ± 5.0 | 82.0 ± 8.7 | 78.5 ± 4.0 |
| 380 | 89.3 ± 5.3 | 88.5 ± 4.5 | 80.0 ± 8.3 | 76.2 ± 3.9 |
| 390 | 88.7 ± 4.0 | 87.5 ± 4.3 | 78.9 ± 7.2 | 74.7 ± 2.5 |
| 400 | 88.9 ± 3.6 | 87.5 ± 3.5 | 77.2 ± 5.5 | 75.0 ± 2.3 |