s5_e0
done · peak unseen_mismatch 92.7
α schedule
α 5.0 ×50 α 4.7826 ×50 α 4.5652 ×50 α 4.3478 ×50 α 4.1304 ×50 α 3.913 ×50 α 3.6957 ×50 α 3.4783 ×50 α 3.2609 ×50 α 3.0435 ×50 α 2.8261 ×50 α 2.6087 ×50 α 2.3913 ×50 α 2.1739 ×50 α 1.9565 ×50 α 1.7391 ×50 α 1.5217 ×50 α 1.3043 ×50 α 1.087 ×50 α 0.8696 ×50 α 0.6522 ×50 α 0.4348 ×50 α 0.2174 ×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 | 51.8 ± 2.1 | 49.9 ± 1.2 | 50.2 ± 1.8 | 50.0 ± 1.4 |
| 100 | 60.0 ± 7.6 | 58.3 ± 7.1 | 53.8 ± 4.8 | 52.9 ± 4.3 |
| 150 | 76.5 ± 6.1 | 74.7 ± 5.8 | 61.8 ± 4.1 | 59.2 ± 4.4 |
| 200 | 80.6 ± 1.8 | 79.1 ± 1.6 | 63.2 ± 3.6 | 61.2 ± 3.0 |
| 250 | 83.5 ± 2.0 | 82.4 ± 2.0 | 65.2 ± 4.4 | 63.4 ± 3.8 |
| 300 | 85.4 ± 1.6 | 83.8 ± 2.0 | 66.7 ± 3.0 | 64.9 ± 2.2 |
| 350 | 88.7 ± 2.5 | 86.7 ± 3.0 | 70.1 ± 5.0 | 68.2 ± 6.0 |
| 400 | 88.1 ± 2.0 | 86.3 ± 1.7 | 68.4 ± 4.6 | 65.7 ± 4.6 |
| 450 | 91.2 ± 1.2 | 89.2 ± 1.6 | 73.0 ± 5.0 | 69.3 ± 5.8 |
| 500 | 91.9 ± 1.5 | 89.3 ± 1.6 | 74.9 ± 5.0 | 69.3 ± 4.7 |
| 550 | 92.9 ± 1.4 | 90.5 ± 1.8 | 77.6 ± 4.7 | 72.3 ± 6.1 |
| 600 | 94.0 ± 0.9 | 91.8 ± 2.2 | 78.1 ± 4.5 | 71.1 ± 5.9 |
| 650 | 95.3 ± 1.2 | 92.7 ± 1.6 | 82.2 ± 6.7 | 74.4 ± 6.4 |
| 700 | 96.2 ± 0.6 | 94.3 ± 1.8 | 83.7 ± 3.5 | 75.7 ± 7.3 |
| 750 | 97.9 ± 0.6 | 95.7 ± 1.5 | 89.0 ± 3.0 | 78.6 ± 6.8 |
| 800 | 98.6 ± 0.6 | 96.7 ± 1.8 | 91.3 ± 4.3 | 82.2 ± 9.2 |
| 850 | 99.0 ± 0.5 | 97.4 ± 1.4 | 94.0 ± 2.7 | 84.0 ± 6.9 |
| 900 | 99.8 ± 0.2 | 98.8 ± 1.3 | 97.8 ± 1.5 | 88.0 ± 8.7 |
| 950 | 99.9 ± 0.2 | 99.3 ± 0.8 | 98.9 ± 0.7 | 91.1 ± 8.2 |
| 1000 | 100.0 ± 0.0 | 99.0 ± 1.1 | 99.4 ± 0.5 | 90.0 ± 10.3 |
| 1050 | 100.0 ± 0.0 | 98.0 ± 3.1 | 99.9 ± 0.1 | 92.7 ± 8.6 |
| 1100 | 100.0 ± 0.0 | 85.8 ± 22.1 | 99.9 ± 0.2 | 78.8 ± 24.4 |
| 1150 | 100.0 ± 0.0 | 56.1 ± 33.6 | 99.9 ± 0.1 | 49.8 ± 29.0 |
| 1200 | 100.0 ± 0.0 | 25.0 ± 25.3 | 100.0 ± 0.1 | 25.2 ± 20.8 |