s5_e3
done · peak unseen_mismatch 75.4
α schedule
α 5.0 ×50 α 4.913 ×50 α 4.8261 ×50 α 4.7391 ×50 α 4.6522 ×50 α 4.5652 ×50 α 4.4783 ×50 α 4.3913 ×50 α 4.3043 ×50 α 4.2174 ×50 α 4.1304 ×50 α 4.0435 ×50 α 3.9565 ×50 α 3.8696 ×50 α 3.7826 ×50 α 3.6957 ×50 α 3.6087 ×50 α 3.5217 ×50 α 3.4348 ×50 α 3.3478 ×50 α 3.2609 ×50 α 3.1739 ×50 α 3.087 ×50 α 3.0 ×50
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 3.0
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 | 61.7 ± 6.8 | 58.1 ± 5.2 | 54.0 ± 3.7 | 51.3 ± 2.1 |
| 150 | 79.2 ± 3.5 | 77.5 ± 2.6 | 65.3 ± 5.7 | 62.7 ± 5.4 |
| 200 | 80.3 ± 2.9 | 79.6 ± 2.9 | 64.0 ± 4.0 | 61.6 ± 3.9 |
| 250 | 81.2 ± 3.3 | 80.6 ± 2.9 | 64.5 ± 3.7 | 62.2 ± 4.7 |
| 300 | 83.4 ± 3.5 | 82.8 ± 3.2 | 66.8 ± 6.5 | 64.8 ± 6.4 |
| 350 | 84.0 ± 2.3 | 82.9 ± 1.9 | 65.5 ± 3.9 | 63.6 ± 4.1 |
| 400 | 86.0 ± 1.8 | 85.4 ± 2.1 | 69.2 ± 3.8 | 66.3 ± 4.5 |
| 450 | 85.2 ± 1.9 | 84.6 ± 2.2 | 67.2 ± 3.4 | 63.7 ± 4.4 |
| 500 | 86.6 ± 2.7 | 85.0 ± 3.2 | 68.0 ± 4.6 | 64.1 ± 3.9 |
| 550 | 86.9 ± 2.2 | 85.7 ± 2.4 | 68.6 ± 3.7 | 64.3 ± 3.5 |
| 600 | 87.8 ± 2.3 | 85.7 ± 2.0 | 69.5 ± 4.4 | 65.8 ± 4.6 |
| 650 | 88.4 ± 1.4 | 86.4 ± 1.8 | 69.8 ± 3.5 | 65.0 ± 3.5 |
| 700 | 88.9 ± 2.4 | 87.0 ± 1.9 | 72.0 ± 3.8 | 67.0 ± 4.1 |
| 750 | 89.2 ± 2.0 | 88.0 ± 2.0 | 72.8 ± 3.4 | 68.2 ± 3.2 |
| 800 | 89.7 ± 2.1 | 88.5 ± 2.3 | 73.4 ± 4.1 | 69.6 ± 4.0 |
| 850 | 90.7 ± 1.2 | 88.7 ± 1.2 | 73.7 ± 4.3 | 68.0 ± 4.3 |
| 900 | 90.6 ± 1.7 | 89.0 ± 1.6 | 73.2 ± 3.5 | 69.3 ± 3.2 |
| 950 | 91.2 ± 1.8 | 89.7 ± 2.0 | 74.7 ± 4.8 | 70.7 ± 4.3 |
| 1000 | 91.8 ± 1.0 | 90.3 ± 1.1 | 75.7 ± 2.8 | 72.2 ± 3.3 |
| 1050 | 92.3 ± 1.0 | 90.5 ± 1.8 | 75.2 ± 3.6 | 70.6 ± 5.4 |
| 1100 | 92.7 ± 1.2 | 91.1 ± 1.1 | 77.3 ± 4.6 | 72.2 ± 4.2 |
| 1150 | 93.0 ± 1.3 | 91.9 ± 1.7 | 79.4 ± 3.4 | 74.2 ± 5.4 |
| 1200 | 94.4 ± 0.6 | 92.3 ± 1.4 | 81.5 ± 2.7 | 75.4 ± 3.7 |