s4_e1
done · peak unseen_mismatch 93.7
α schedule
α 4.0 ×50 α 3.8696 ×50 α 3.7391 ×50 α 3.6087 ×50 α 3.4783 ×50 α 3.3478 ×50 α 3.2174 ×50 α 3.087 ×50 α 2.9565 ×50 α 2.8261 ×50 α 2.6957 ×50 α 2.5652 ×50 α 2.4348 ×50 α 2.3043 ×50 α 2.1739 ×50 α 2.0435 ×50 α 1.913 ×50 α 1.7826 ×50 α 1.6522 ×50 α 1.5217 ×50 α 1.3913 ×50 α 1.2609 ×50 α 1.1304 ×50 α 1.0 ×50
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.0
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 | 51.7 ± 1.9 | 49.6 ± 1.0 | 49.9 ± 2.1 | 50.5 ± 1.2 |
| 100 | 58.5 ± 8.5 | 56.3 ± 6.7 | 54.2 ± 5.7 | 53.0 ± 4.2 |
| 150 | 77.8 ± 6.9 | 76.9 ± 6.2 | 65.6 ± 7.0 | 62.0 ± 6.6 |
| 200 | 83.7 ± 2.0 | 81.9 ± 1.9 | 67.7 ± 3.7 | 64.5 ± 3.5 |
| 250 | 86.7 ± 2.1 | 84.7 ± 1.9 | 69.2 ± 5.3 | 67.3 ± 3.4 |
| 300 | 89.5 ± 1.9 | 87.9 ± 1.7 | 73.0 ± 3.3 | 70.3 ± 4.1 |
| 350 | 90.2 ± 2.3 | 88.7 ± 2.6 | 74.0 ± 5.6 | 70.7 ± 4.5 |
| 400 | 91.0 ± 2.1 | 88.9 ± 1.9 | 74.2 ± 5.5 | 70.0 ± 4.6 |
| 450 | 92.7 ± 1.0 | 90.5 ± 1.7 | 77.8 ± 4.6 | 72.3 ± 4.6 |
| 500 | 93.4 ± 1.1 | 91.9 ± 1.0 | 77.8 ± 2.3 | 73.0 ± 2.6 |
| 550 | 94.2 ± 1.0 | 92.2 ± 1.2 | 80.3 ± 3.1 | 75.7 ± 3.7 |
| 600 | 95.0 ± 1.0 | 93.0 ± 1.6 | 82.1 ± 3.7 | 75.9 ± 5.2 |
| 650 | 95.9 ± 1.1 | 93.9 ± 1.0 | 83.2 ± 4.0 | 77.7 ± 4.0 |
| 700 | 96.4 ± 0.7 | 94.7 ± 1.2 | 84.6 ± 2.3 | 78.4 ± 4.1 |
| 750 | 97.2 ± 0.7 | 94.9 ± 1.5 | 87.2 ± 3.4 | 79.0 ± 5.6 |
| 800 | 97.7 ± 1.0 | 96.0 ± 1.5 | 88.2 ± 4.2 | 81.6 ± 5.6 |
| 850 | 98.1 ± 0.4 | 96.7 ± 1.1 | 91.2 ± 2.3 | 84.6 ± 4.1 |
| 900 | 99.0 ± 0.3 | 97.8 ± 1.0 | 93.3 ± 1.5 | 86.1 ± 4.8 |
| 950 | 99.2 ± 0.5 | 98.0 ± 0.8 | 94.3 ± 3.2 | 85.8 ± 6.4 |
| 1000 | 99.5 ± 0.3 | 98.5 ± 0.6 | 96.0 ± 1.9 | 89.2 ± 3.9 |
| 1050 | 99.8 ± 0.2 | 99.2 ± 0.9 | 97.4 ± 1.1 | 89.7 ± 6.5 |
| 1100 | 99.9 ± 0.1 | 99.0 ± 0.8 | 97.6 ± 1.2 | 90.9 ± 4.1 |
| 1150 | 100.0 ± 0.0 | 99.6 ± 0.4 | 99.5 ± 0.3 | 93.7 ± 4.7 |
| 1200 | 100.0 ± 0.0 | 99.3 ± 1.1 | 99.5 ± 0.4 | 91.6 ± 9.3 |