s4_e0
done · peak unseen_mismatch 87.8
α schedule
α 4.0 ×50 α 3.8261 ×50 α 3.6522 ×50 α 3.4783 ×50 α 3.3043 ×50 α 3.1304 ×50 α 2.9565 ×50 α 2.7826 ×50 α 2.6087 ×50 α 2.4348 ×50 α 2.2609 ×50 α 2.087 ×50 α 1.913 ×50 α 1.7391 ×50 α 1.5652 ×50 α 1.3913 ×50 α 1.2174 ×50 α 1.0435 ×50 α 0.8696 ×50 α 0.6957 ×50 α 0.5217 ×50 α 0.3478 ×50 α 0.1739 ×50 α 0.0 ×50
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.0
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 | 56.4 ± 4.8 | 54.0 ± 4.9 | 52.6 ± 2.8 | 50.5 ± 2.1 |
| 150 | 81.9 ± 6.0 | 77.8 ± 6.8 | 71.1 ± 8.1 | 65.5 ± 7.1 |
| 200 | 85.7 ± 3.6 | 84.5 ± 3.8 | 71.4 ± 7.0 | 68.2 ± 6.8 |
| 250 | 86.7 ± 2.0 | 86.4 ± 2.5 | 71.2 ± 4.5 | 68.0 ± 4.9 |
| 300 | 89.4 ± 2.3 | 89.2 ± 2.0 | 75.2 ± 5.1 | 72.9 ± 5.4 |
| 350 | 90.9 ± 1.5 | 90.0 ± 1.4 | 75.7 ± 4.7 | 72.5 ± 4.6 |
| 400 | 92.8 ± 1.6 | 91.3 ± 1.9 | 79.8 ± 5.5 | 75.1 ± 6.0 |
| 450 | 93.3 ± 1.6 | 91.6 ± 1.4 | 79.0 ± 5.7 | 73.7 ± 5.8 |
| 500 | 94.2 ± 1.6 | 92.5 ± 1.4 | 81.4 ± 4.7 | 74.7 ± 6.2 |
| 550 | 95.4 ± 1.8 | 93.8 ± 1.4 | 83.9 ± 4.1 | 77.8 ± 4.9 |
| 600 | 97.2 ± 0.6 | 95.2 ± 1.1 | 87.4 ± 3.0 | 80.8 ± 5.2 |
| 650 | 97.7 ± 0.9 | 96.0 ± 1.0 | 88.8 ± 2.8 | 81.9 ± 3.9 |
| 700 | 98.8 ± 0.5 | 96.6 ± 1.3 | 92.4 ± 2.4 | 83.7 ± 6.4 |
| 750 | 99.4 ± 0.4 | 97.7 ± 0.8 | 95.7 ± 1.5 | 86.5 ± 6.9 |
| 800 | 99.6 ± 0.3 | 98.2 ± 0.8 | 97.3 ± 1.2 | 87.0 ± 7.6 |
| 850 | 99.9 ± 0.1 | 98.2 ± 2.5 | 98.5 ± 1.4 | 87.8 ± 12.8 |
| 900 | 99.9 ± 0.1 | 98.4 ± 2.9 | 98.8 ± 1.0 | 87.0 ± 12.6 |
| 950 | 100.0 ± 0.0 | 96.5 ± 8.4 | 99.8 ± 0.5 | 84.6 ± 19.0 |
| 1000 | 100.0 ± 0.0 | 94.9 ± 8.1 | 99.8 ± 0.3 | 81.1 ± 18.1 |
| 1050 | 100.0 ± 0.0 | 81.6 ± 15.9 | 100.0 ± 0.1 | 62.0 ± 28.1 |
| 1100 | 100.0 ± 0.0 | 60.4 ± 26.8 | 99.9 ± 0.2 | 47.8 ± 24.5 |
| 1150 | 100.0 ± 0.0 | 33.9 ± 23.4 | 100.0 ± 0.1 | 32.6 ± 22.8 |
| 1200 | 100.0 ± 0.0 | 26.1 ± 24.3 | 99.9 ± 0.1 | 26.1 ± 19.8 |