s4_e2
done · peak unseen_mismatch 80.1
α schedule
α 4.0 ×50 α 3.913 ×50 α 3.8261 ×50 α 3.7391 ×50 α 3.6522 ×50 α 3.5652 ×50 α 3.4783 ×50 α 3.3913 ×50 α 3.3043 ×50 α 3.2174 ×50 α 3.1304 ×50 α 3.0435 ×50 α 2.9565 ×50 α 2.8696 ×50 α 2.7826 ×50 α 2.6957 ×50 α 2.6087 ×50 α 2.5217 ×50 α 2.4348 ×50 α 2.3478 ×50 α 2.2609 ×50 α 2.1739 ×50 α 2.087 ×50 α 2.0 ×50
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 2.0
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 | 55.9 ± 4.6 | 53.8 ± 4.2 | 52.2 ± 2.7 | 50.5 ± 2.1 |
| 150 | 81.7 ± 5.5 | 78.0 ± 6.3 | 70.5 ± 8.5 | 65.3 ± 7.3 |
| 200 | 84.6 ± 3.6 | 83.3 ± 3.5 | 69.9 ± 7.1 | 67.0 ± 6.0 |
| 250 | 85.1 ± 1.9 | 84.7 ± 2.1 | 70.0 ± 4.9 | 66.8 ± 4.9 |
| 300 | 87.3 ± 1.7 | 87.2 ± 1.9 | 72.3 ± 4.4 | 69.5 ± 5.4 |
| 350 | 89.0 ± 1.5 | 87.8 ± 1.4 | 73.2 ± 4.2 | 69.9 ± 4.5 |
| 400 | 90.3 ± 1.9 | 89.0 ± 2.4 | 75.9 ± 6.1 | 72.1 ± 7.4 |
| 450 | 90.4 ± 1.8 | 89.0 ± 2.0 | 74.9 ± 4.6 | 70.3 ± 6.0 |
| 500 | 91.6 ± 2.0 | 89.4 ± 2.4 | 76.3 ± 4.6 | 71.1 ± 6.2 |
| 550 | 91.5 ± 1.4 | 89.6 ± 1.4 | 76.3 ± 4.4 | 70.3 ± 4.0 |
| 600 | 92.8 ± 1.6 | 90.8 ± 1.7 | 79.0 ± 4.6 | 72.2 ± 4.9 |
| 650 | 93.1 ± 1.8 | 90.9 ± 1.8 | 78.6 ± 5.2 | 70.7 ± 5.6 |
| 700 | 93.3 ± 1.4 | 91.4 ± 1.2 | 81.5 ± 3.5 | 73.6 ± 3.9 |
| 750 | 94.1 ± 1.2 | 91.4 ± 1.0 | 81.8 ± 3.8 | 73.7 ± 4.5 |
| 800 | 94.5 ± 1.1 | 92.0 ± 1.3 | 82.1 ± 5.0 | 74.2 ± 4.9 |
| 850 | 95.0 ± 0.9 | 93.0 ± 1.3 | 83.6 ± 3.4 | 74.1 ± 5.0 |
| 900 | 95.7 ± 1.1 | 93.3 ± 1.2 | 85.7 ± 3.6 | 76.2 ± 5.1 |
| 950 | 96.5 ± 1.0 | 93.6 ± 1.5 | 87.0 ± 3.8 | 76.5 ± 5.6 |
| 1000 | 96.7 ± 1.0 | 94.4 ± 0.8 | 87.8 ± 1.9 | 78.3 ± 4.8 |
| 1050 | 97.3 ± 0.7 | 94.5 ± 1.3 | 87.2 ± 3.0 | 76.7 ± 6.0 |
| 1100 | 97.6 ± 0.6 | 95.5 ± 0.9 | 88.9 ± 3.1 | 79.6 ± 4.7 |
| 1150 | 97.8 ± 0.9 | 95.2 ± 1.0 | 89.9 ± 2.4 | 78.9 ± 5.9 |
| 1200 | 98.3 ± 0.6 | 96.2 ± 1.2 | 91.6 ± 2.6 | 80.1 ± 6.7 |