s5_e5
done · peak unseen_mismatch 66.0
α schedule
α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50 α 5.0 ×50
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 5.0
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 |
|---|---|---|---|---|
| 50 | 51.3 ± 1.5 | 50.3 ± 1.3 | 50.3 ± 1.5 | 50.3 ± 1.8 |
| 100 | 62.0 ± 6.5 | 58.2 ± 5.2 | 54.0 ± 3.7 | 51.2 ± 2.1 |
| 150 | 77.9 ± 3.6 | 75.9 ± 3.0 | 64.6 ± 4.8 | 61.5 ± 4.3 |
| 200 | 80.0 ± 2.7 | 78.8 ± 2.9 | 62.8 ± 3.7 | 61.2 ± 3.6 |
| 250 | 80.2 ± 3.2 | 79.8 ± 2.8 | 64.4 ± 3.9 | 62.2 ± 4.1 |
| 300 | 81.6 ± 3.4 | 80.9 ± 2.8 | 64.5 ± 6.1 | 63.1 ± 5.3 |
| 350 | 81.7 ± 2.1 | 80.9 ± 1.4 | 63.0 ± 3.3 | 61.0 ± 3.5 |
| 400 | 83.7 ± 2.0 | 83.5 ± 2.0 | 67.1 ± 4.3 | 64.8 ± 4.5 |
| 450 | 82.6 ± 1.9 | 82.2 ± 2.1 | 64.8 ± 4.2 | 62.1 ± 4.7 |
| 500 | 84.1 ± 3.1 | 83.0 ± 3.0 | 66.6 ± 6.3 | 63.4 ± 6.0 |
| 550 | 84.0 ± 2.0 | 82.4 ± 2.3 | 65.2 ± 3.0 | 61.8 ± 3.4 |
| 600 | 85.1 ± 2.4 | 83.4 ± 2.4 | 68.1 ± 4.9 | 64.4 ± 4.6 |
| 650 | 84.9 ± 1.5 | 83.3 ± 1.7 | 66.8 ± 3.8 | 63.2 ± 3.7 |
| 700 | 85.7 ± 1.8 | 83.4 ± 2.2 | 67.6 ± 3.7 | 63.5 ± 3.7 |
| 750 | 85.9 ± 1.7 | 84.4 ± 1.5 | 68.8 ± 3.6 | 65.1 ± 3.4 |
| 800 | 85.4 ± 2.8 | 84.1 ± 2.9 | 67.8 ± 4.4 | 63.6 ± 5.0 |
| 850 | 86.2 ± 1.7 | 85.1 ± 2.1 | 68.9 ± 3.8 | 64.2 ± 4.8 |
| 900 | 85.9 ± 1.5 | 84.6 ± 1.8 | 67.9 ± 3.5 | 63.9 ± 4.6 |
| 950 | 86.6 ± 1.3 | 85.1 ± 1.6 | 69.0 ± 4.4 | 65.1 ± 4.9 |
| 1000 | 86.6 ± 1.7 | 85.1 ± 2.1 | 69.3 ± 3.8 | 65.1 ± 5.1 |
| 1050 | 86.5 ± 1.8 | 85.4 ± 1.5 | 69.6 ± 3.1 | 65.3 ± 4.2 |
| 1100 | 87.1 ± 1.9 | 85.3 ± 1.8 | 68.9 ± 2.8 | 64.8 ± 3.7 |
| 1150 | 87.7 ± 1.6 | 86.1 ± 1.9 | 69.1 ± 3.8 | 65.6 ± 4.7 |
| 1200 | 87.5 ± 1.1 | 86.0 ± 1.6 | 69.2 ± 3.2 | 66.0 ± 3.9 |