r3_c5
done · peak unseen_mismatch 98.4
α schedule
α 3.2007 ×25 α 3.1661 ×25 α 3.1038 ×25 α 3.016 ×25 α 2.904 ×25 α 2.7689 ×25 α 2.6113 ×25 α 2.4318 ×25 α 2.2309 ×25 α 2.009 ×25 α 1.7666 ×25 α 1.5038 ×25 α 1.221 ×25 α 0.9185 ×25 α 0.5965 ×25 α 0.2551 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.2551
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 | 50.3 ± 1.8 | 50.0 ± 1.6 | 49.5 ± 1.7 |
| 50 | 51.6 ± 1.5 | 50.3 ± 1.7 | 50.1 ± 1.3 | 49.8 ± 1.8 |
| 75 | 52.1 ± 1.7 | 49.9 ± 1.7 | 50.5 ± 1.5 | 49.4 ± 1.6 |
| 100 | 56.1 ± 7.3 | 53.3 ± 4.0 | 52.7 ± 5.2 | 50.2 ± 2.0 |
| 125 | 67.2 ± 13.2 | 62.5 ± 11.3 | 58.9 ± 11.6 | 55.4 ± 9.1 |
| 150 | 77.4 ± 12.7 | 74.0 ± 12.3 | 68.7 ± 13.3 | 64.6 ± 12.0 |
| 175 | 85.6 ± 8.6 | 82.5 ± 8.2 | 74.6 ± 11.8 | 70.3 ± 10.9 |
| 200 | 91.6 ± 4.9 | 89.0 ± 4.9 | 80.9 ± 9.6 | 77.3 ± 10.0 |
| 225 | 93.7 ± 2.8 | 91.5 ± 3.3 | 83.6 ± 7.0 | 79.8 ± 7.7 |
| 250 | 95.1 ± 1.7 | 93.5 ± 2.1 | 85.6 ± 5.3 | 82.1 ± 5.7 |
| 275 | 97.3 ± 1.1 | 95.8 ± 1.4 | 90.1 ± 4.5 | 86.5 ± 4.8 |
| 300 | 98.6 ± 0.9 | 97.4 ± 1.4 | 94.0 ± 3.4 | 90.6 ± 4.3 |
| 325 | 99.5 ± 0.5 | 98.6 ± 0.9 | 96.8 ± 2.8 | 93.6 ± 4.8 |
| 350 | 100.0 ± 0.1 | 99.6 ± 0.6 | 99.6 ± 0.5 | 97.6 ± 2.7 |
| 375 | 100.0 ± 0.0 | 99.8 ± 0.5 | 100.0 ± 0.1 | 98.4 ± 2.1 |
| 400 | 100.0 ± 0.0 | 94.0 ± 14.1 | 100.0 ± 0.0 | 91.5 ± 15.8 |