r3_c0
done · peak unseen_mismatch 90.6
α schedule
α 1.868 ×25 α 1.8736 ×25 α 1.8862 ×25 α 1.9064 ×25 α 1.9346 ×25 α 1.9711 ×25 α 2.0163 ×25 α 2.0704 ×25 α 2.1335 ×25 α 2.2059 ×25 α 2.2877 ×25 α 2.3792 ×25 α 2.4804 ×25 α 2.5914 ×25 α 2.7125 ×25 α 2.8437 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 2.8437
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 | 50.5 ± 1.4 | 50.1 ± 1.5 | 50.1 ± 1.5 | 49.4 ± 1.8 |
| 50 | 50.9 ± 1.9 | 50.2 ± 1.3 | 50.1 ± 1.3 | 50.1 ± 1.3 |
| 75 | 50.9 ± 1.8 | 49.6 ± 1.6 | 50.4 ± 1.5 | 49.5 ± 1.7 |
| 100 | 50.8 ± 2.4 | 50.2 ± 1.4 | 50.1 ± 1.3 | 50.1 ± 1.6 |
| 125 | 52.4 ± 2.2 | 49.6 ± 1.6 | 51.1 ± 2.2 | 49.2 ± 1.8 |
| 150 | 59.6 ± 12.6 | 52.2 ± 5.0 | 57.1 ± 11.7 | 50.3 ± 4.5 |
| 175 | 70.5 ± 18.6 | 60.1 ± 11.1 | 69.2 ± 18.7 | 58.6 ± 10.5 |
| 200 | 79.3 ± 20.7 | 67.1 ± 16.7 | 78.1 ± 20.9 | 65.6 ± 16.2 |
| 225 | 86.5 ± 17.1 | 78.0 ± 16.9 | 85.2 ± 17.6 | 75.9 ± 16.5 |
| 250 | 93.0 ± 12.7 | 87.1 ± 14.3 | 91.1 ± 13.2 | 84.0 ± 14.8 |
| 275 | 95.7 ± 10.2 | 91.7 ± 10.9 | 93.6 ± 11.4 | 88.7 ± 11.3 |
| 300 | 97.3 ± 4.4 | 94.6 ± 5.9 | 94.7 ± 5.9 | 90.6 ± 7.5 |
| 325 | 97.8 ± 1.6 | 95.5 ± 2.9 | 94.6 ± 4.5 | 90.5 ± 6.4 |
| 350 | 97.3 ± 2.1 | 95.3 ± 2.9 | 92.6 ± 6.0 | 88.8 ± 7.4 |
| 375 | 96.3 ± 1.8 | 94.5 ± 2.4 | 90.7 ± 6.0 | 86.9 ± 7.2 |
| 400 | 95.6 ± 1.7 | 93.9 ± 2.1 | 89.2 ± 5.2 | 85.6 ± 5.9 |