free-dcce7e2c
done · peak unseen_mismatch 98.0
α schedule
α 4.486 ×47 α 2.385 ×42 α 4.197 ×31 α 0.4647 ×21 α 0.4635 ×13 α 3.347 ×20 α 2.643 ×34 α 3.413 ×69 α 0.2656 ×23 α 0.2656 ×100
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.2656
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 | 57.4 ± 7.7 | 53.6 ± 5.7 | 53.1 ± 6.0 | 50.6 ± 4.2 |
| 110 | 61.8 ± 8.8 | 57.2 ± 7.1 | 55.5 ± 7.1 | 51.9 ± 5.4 |
| 120 | 67.4 ± 10.4 | 62.1 ± 8.6 | 59.3 ± 9.1 | 54.9 ± 6.7 |
| 130 | 70.1 ± 12.2 | 63.5 ± 10.3 | 63.7 ± 12.5 | 57.9 ± 9.1 |
| 140 | 76.5 ± 15.6 | 66.0 ± 12.1 | 72.5 ± 16.8 | 61.6 ± 11.9 |
| 150 | 83.9 ± 17.6 | 65.2 ± 14.4 | 81.5 ± 19.0 | 62.6 ± 14.2 |
| 160 | 88.9 ± 16.5 | 67.3 ± 16.9 | 87.4 ± 18.0 | 64.8 ± 16.6 |
| 170 | 91.7 ± 14.4 | 76.8 ± 16.3 | 90.0 ± 16.3 | 74.3 ± 15.9 |
| 180 | 92.9 ± 11.8 | 82.0 ± 13.6 | 90.8 ± 14.3 | 78.8 ± 14.5 |
| 190 | 93.6 ± 10.2 | 83.6 ± 12.1 | 90.7 ± 12.9 | 79.8 ± 12.3 |
| 200 | 94.0 ± 8.3 | 85.8 ± 11.2 | 90.6 ± 11.1 | 81.4 ± 11.4 |
| 210 | 94.5 ± 6.6 | 89.1 ± 9.4 | 90.5 ± 9.4 | 84.1 ± 9.7 |
| 220 | 93.5 ± 3.9 | 88.3 ± 7.1 | 87.6 ± 7.5 | 80.8 ± 7.7 |
| 230 | 93.0 ± 3.1 | 88.0 ± 6.0 | 86.1 ± 7.6 | 79.4 ± 6.0 |
| 240 | 92.9 ± 3.6 | 89.9 ± 3.7 | 85.3 ± 7.2 | 80.4 ± 6.9 |
| 250 | 92.1 ± 3.3 | 89.2 ± 3.3 | 83.4 ± 7.0 | 79.1 ± 5.8 |
| 260 | 92.0 ± 2.9 | 88.7 ± 2.9 | 82.8 ± 7.4 | 78.4 ± 5.7 |
| 270 | 91.4 ± 2.9 | 88.4 ± 3.0 | 80.7 ± 7.2 | 76.8 ± 6.3 |
| 280 | 93.2 ± 2.7 | 90.6 ± 2.7 | 84.2 ± 6.8 | 80.6 ± 5.3 |
| 290 | 98.7 ± 1.6 | 97.8 ± 2.2 | 97.1 ± 3.7 | 95.5 ± 4.5 |
| 300 | 99.7 ± 0.1 | 98.9 ± 0.8 | 98.9 ± 0.0 | 98.0 ± 0.6 |