r3_c4
done · peak unseen_mismatch 97.7
α schedule
α 3.5999 ×25 α 3.5832 ×25 α 3.5451 ×25 α 3.4832 ×25 α 3.3957 ×25 α 3.2812 ×25 α 3.1388 ×25 α 2.9672 ×25 α 2.7658 ×25 α 2.5338 ×25 α 2.2704 ×25 α 1.975 ×25 α 1.6469 ×25 α 1.2858 ×25 α 0.8909 ×25 α 0.4619 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.4619
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 |
|---|---|---|---|---|
| 25 | 50.9 ± 1.4 | 50.4 ± 1.6 | 49.9 ± 1.4 | 49.5 ± 1.9 |
| 50 | 51.2 ± 1.4 | 50.3 ± 1.6 | 49.9 ± 1.1 | 50.0 ± 1.6 |
| 75 | 52.3 ± 2.2 | 50.0 ± 1.6 | 50.6 ± 1.5 | 49.4 ± 1.4 |
| 100 | 57.0 ± 6.9 | 54.9 ± 4.3 | 51.9 ± 3.6 | 50.2 ± 1.7 |
| 125 | 68.7 ± 11.2 | 65.1 ± 10.0 | 59.1 ± 9.9 | 56.5 ± 7.5 |
| 150 | 79.2 ± 8.4 | 76.4 ± 8.3 | 66.7 ± 9.6 | 63.4 ± 8.7 |
| 175 | 85.4 ± 4.2 | 82.5 ± 4.2 | 71.3 ± 9.5 | 68.1 ± 9.1 |
| 200 | 87.2 ± 3.6 | 84.9 ± 3.6 | 71.5 ± 7.5 | 69.1 ± 8.0 |
| 225 | 89.0 ± 2.9 | 86.6 ± 2.7 | 73.6 ± 6.6 | 71.1 ± 6.8 |
| 250 | 90.7 ± 2.2 | 88.8 ± 2.0 | 75.4 ± 5.9 | 72.7 ± 5.9 |
| 275 | 93.2 ± 1.8 | 91.5 ± 1.6 | 79.0 ± 5.4 | 76.8 ± 4.8 |
| 300 | 95.4 ± 1.5 | 93.8 ± 1.5 | 83.7 ± 5.0 | 80.5 ± 4.9 |
| 325 | 97.3 ± 0.9 | 96.2 ± 1.2 | 88.8 ± 4.6 | 85.3 ± 5.6 |
| 350 | 99.5 ± 0.4 | 98.9 ± 0.4 | 96.6 ± 1.7 | 93.9 ± 2.3 |
| 375 | 99.9 ± 0.1 | 99.8 ± 0.2 | 99.2 ± 0.6 | 97.7 ± 1.5 |
| 400 | 100.0 ± 0.0 | 97.2 ± 5.1 | 100.0 ± 0.1 | 94.5 ± 8.6 |