s5_e5
done · peak unseen_mismatch 66.2
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 | 51.8 ± 2.1 | 49.9 ± 1.2 | 50.2 ± 1.8 | 50.0 ± 1.4 |
| 100 | 60.2 ± 7.1 | 58.2 ± 6.8 | 53.9 ± 4.6 | 52.4 ± 4.1 |
| 150 | 75.1 ± 5.2 | 74.1 ± 5.3 | 60.7 ± 3.6 | 58.2 ± 3.8 |
| 200 | 79.5 ± 1.8 | 77.8 ± 1.4 | 62.5 ± 3.3 | 60.9 ± 3.3 |
| 250 | 80.9 ± 2.3 | 79.8 ± 1.4 | 62.2 ± 3.2 | 60.6 ± 2.4 |
| 300 | 82.3 ± 2.3 | 80.8 ± 2.9 | 64.5 ± 3.8 | 63.1 ± 3.0 |
| 350 | 84.2 ± 2.3 | 82.5 ± 2.5 | 65.1 ± 3.9 | 63.3 ± 3.4 |
| 400 | 83.3 ± 1.5 | 81.5 ± 1.9 | 64.3 ± 1.4 | 61.6 ± 2.5 |
| 450 | 84.5 ± 2.0 | 82.0 ± 2.3 | 64.1 ± 4.0 | 60.4 ± 4.1 |
| 500 | 84.5 ± 1.8 | 82.3 ± 2.1 | 66.0 ± 4.3 | 62.0 ± 5.0 |
| 550 | 85.0 ± 2.0 | 82.9 ± 2.8 | 66.7 ± 6.5 | 63.3 ± 7.4 |
| 600 | 85.3 ± 1.8 | 83.4 ± 1.8 | 67.3 ± 3.3 | 62.9 ± 3.9 |
| 650 | 86.1 ± 1.5 | 84.0 ± 1.6 | 67.2 ± 4.6 | 63.2 ± 5.4 |
| 700 | 85.7 ± 1.7 | 83.6 ± 1.4 | 66.0 ± 4.0 | 63.0 ± 3.8 |
| 750 | 86.3 ± 1.3 | 84.0 ± 1.6 | 68.2 ± 2.9 | 63.2 ± 3.4 |
| 800 | 85.9 ± 1.9 | 83.5 ± 1.6 | 67.3 ± 3.3 | 63.8 ± 3.5 |
| 850 | 86.8 ± 1.0 | 84.2 ± 1.8 | 68.3 ± 5.1 | 63.8 ± 6.0 |
| 900 | 86.2 ± 1.4 | 83.5 ± 1.4 | 67.9 ± 4.0 | 62.6 ± 4.6 |
| 950 | 87.2 ± 1.4 | 84.5 ± 1.6 | 69.8 ± 3.9 | 64.2 ± 4.7 |
| 1000 | 86.8 ± 1.2 | 84.7 ± 1.7 | 68.1 ± 3.5 | 63.6 ± 3.6 |
| 1050 | 87.8 ± 1.0 | 85.1 ± 1.6 | 69.4 ± 3.2 | 64.9 ± 4.5 |
| 1100 | 87.3 ± 1.5 | 84.8 ± 1.9 | 69.0 ± 3.9 | 64.2 ± 3.8 |
| 1150 | 87.6 ± 1.3 | 85.5 ± 1.4 | 69.3 ± 3.6 | 65.7 ± 4.3 |
| 1200 | 87.8 ± 1.3 | 85.8 ± 0.9 | 70.5 ± 3.2 | 66.2 ± 3.9 |