s4_e3
done · peak unseen_mismatch 75.9
α schedule
α 4.0 ×50 α 3.9565 ×50 α 3.913 ×50 α 3.8696 ×50 α 3.8261 ×50 α 3.7826 ×50 α 3.7391 ×50 α 3.6957 ×50 α 3.6522 ×50 α 3.6087 ×50 α 3.5652 ×50 α 3.5217 ×50 α 3.4783 ×50 α 3.4348 ×50 α 3.3913 ×50 α 3.3478 ×50 α 3.3043 ×50 α 3.2609 ×50 α 3.2174 ×50 α 3.1739 ×50 α 3.1304 ×50 α 3.087 ×50 α 3.0435 ×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 | 51.4 ± 1.2 | 50.2 ± 1.3 | 49.9 ± 1.6 | 50.4 ± 1.8 |
| 100 | 56.2 ± 4.5 | 53.8 ± 4.3 | 52.2 ± 2.9 | 50.6 ± 2.1 |
| 150 | 80.6 ± 5.4 | 77.5 ± 6.2 | 69.6 ± 7.7 | 64.5 ± 7.1 |
| 200 | 84.2 ± 3.7 | 82.8 ± 3.3 | 69.1 ± 6.1 | 66.6 ± 5.3 |
| 250 | 84.9 ± 2.4 | 84.3 ± 2.4 | 69.5 ± 5.3 | 66.3 ± 5.2 |
| 300 | 87.3 ± 2.1 | 86.8 ± 1.8 | 71.5 ± 3.7 | 70.0 ± 4.7 |
| 350 | 88.1 ± 1.7 | 87.3 ± 1.4 | 71.7 ± 3.2 | 68.8 ± 3.4 |
| 400 | 89.6 ± 1.9 | 88.4 ± 2.4 | 75.1 ± 5.9 | 70.9 ± 6.3 |
| 450 | 89.5 ± 1.6 | 87.7 ± 1.1 | 73.0 ± 3.8 | 68.4 ± 3.9 |
| 500 | 89.9 ± 1.5 | 88.1 ± 2.1 | 73.3 ± 3.9 | 69.4 ± 4.0 |
| 550 | 90.4 ± 1.6 | 89.2 ± 2.5 | 74.6 ± 4.8 | 70.2 ± 3.7 |
| 600 | 90.9 ± 1.7 | 89.3 ± 1.8 | 75.8 ± 4.7 | 71.0 ± 4.9 |
| 650 | 91.0 ± 1.3 | 89.7 ± 1.7 | 77.1 ± 3.6 | 71.8 ± 3.2 |
| 700 | 92.1 ± 1.6 | 89.9 ± 1.4 | 78.2 ± 2.6 | 73.1 ± 3.7 |
| 750 | 91.8 ± 1.2 | 90.4 ± 1.6 | 78.7 ± 3.6 | 73.2 ± 3.6 |
| 800 | 92.0 ± 1.7 | 90.4 ± 1.3 | 78.6 ± 4.3 | 73.3 ± 4.2 |
| 850 | 92.2 ± 1.2 | 91.0 ± 1.7 | 77.9 ± 4.2 | 72.1 ± 4.3 |
| 900 | 92.5 ± 1.6 | 90.7 ± 2.4 | 79.3 ± 2.5 | 72.2 ± 4.8 |
| 950 | 93.1 ± 1.4 | 91.5 ± 2.0 | 79.2 ± 4.3 | 73.5 ± 3.9 |
| 1000 | 93.5 ± 1.3 | 91.6 ± 1.6 | 80.5 ± 2.5 | 74.0 ± 4.1 |
| 1050 | 93.8 ± 1.5 | 92.0 ± 1.8 | 79.5 ± 5.0 | 73.5 ± 5.5 |
| 1100 | 93.5 ± 1.0 | 91.7 ± 1.4 | 79.8 ± 4.1 | 72.5 ± 5.4 |
| 1150 | 94.6 ± 1.4 | 92.8 ± 1.3 | 81.9 ± 3.4 | 75.5 ± 4.3 |
| 1200 | 94.3 ± 1.2 | 92.9 ± 2.0 | 81.5 ± 3.7 | 75.9 ± 4.3 |