s3_e0
done · peak unseen_mismatch 92.4
α schedule
α 3.0 ×50 α 2.8696 ×50 α 2.7391 ×50 α 2.6087 ×50 α 2.4783 ×50 α 2.3478 ×50 α 2.2174 ×50 α 2.087 ×50 α 1.9565 ×50 α 1.8261 ×50 α 1.6957 ×50 α 1.5652 ×50 α 1.4348 ×50 α 1.3043 ×50 α 1.1739 ×50 α 1.0435 ×50 α 0.913 ×50 α 0.7826 ×50 α 0.6522 ×50 α 0.5217 ×50 α 0.3913 ×50 α 0.2609 ×50 α 0.1304 ×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 | 51.3 ± 2.2 | 49.6 ± 0.8 | 50.0 ± 1.7 | 50.3 ± 1.6 |
| 100 | 56.3 ± 6.4 | 53.4 ± 4.4 | 53.5 ± 5.6 | 50.8 ± 2.4 |
| 150 | 76.6 ± 15.2 | 73.8 ± 14.0 | 69.6 ± 15.9 | 65.2 ± 13.5 |
| 200 | 91.5 ± 3.4 | 89.1 ± 3.4 | 82.0 ± 6.0 | 77.3 ± 5.5 |
| 250 | 93.9 ± 1.1 | 92.6 ± 1.3 | 82.8 ± 4.7 | 78.8 ± 4.7 |
| 300 | 94.2 ± 1.4 | 92.8 ± 1.1 | 82.0 ± 4.4 | 78.0 ± 5.1 |
| 350 | 95.2 ± 1.5 | 93.9 ± 1.7 | 83.4 ± 3.5 | 78.5 ± 4.4 |
| 400 | 96.3 ± 1.4 | 94.3 ± 1.6 | 85.7 ± 4.9 | 79.9 ± 4.6 |
| 450 | 97.6 ± 0.9 | 95.1 ± 1.3 | 88.5 ± 3.6 | 80.9 ± 3.0 |
| 500 | 97.9 ± 1.0 | 96.6 ± 1.0 | 90.3 ± 3.0 | 82.6 ± 4.4 |
| 550 | 98.6 ± 0.5 | 96.9 ± 0.8 | 91.9 ± 2.6 | 83.6 ± 3.8 |
| 600 | 99.3 ± 0.5 | 98.3 ± 0.6 | 95.2 ± 2.9 | 87.6 ± 4.1 |
| 650 | 99.6 ± 0.3 | 98.4 ± 0.9 | 96.2 ± 1.5 | 89.1 ± 3.5 |
| 700 | 99.7 ± 0.3 | 98.7 ± 1.2 | 96.8 ± 1.9 | 88.8 ± 6.4 |
| 750 | 100.0 ± 0.1 | 99.3 ± 0.8 | 99.0 ± 0.9 | 92.4 ± 5.5 |
| 800 | 100.0 ± 0.0 | 98.9 ± 1.0 | 99.0 ± 0.6 | 89.0 ± 6.6 |
| 850 | 100.0 ± 0.0 | 98.8 ± 2.4 | 99.7 ± 0.3 | 91.6 ± 9.5 |
| 900 | 100.0 ± 0.0 | 98.4 ± 2.3 | 99.8 ± 0.4 | 90.1 ± 11.1 |
| 950 | 100.0 ± 0.0 | 96.8 ± 2.3 | 99.9 ± 0.2 | 83.6 ± 9.1 |
| 1000 | 100.0 ± 0.0 | 79.4 ± 20.5 | 99.6 ± 0.8 | 67.5 ± 20.9 |
| 1050 | 100.0 ± 0.0 | 59.5 ± 24.1 | 99.8 ± 0.4 | 47.6 ± 23.4 |
| 1100 | 100.0 ± 0.0 | 32.9 ± 14.1 | 99.9 ± 0.1 | 27.5 ± 12.2 |
| 1150 | 100.0 ± 0.0 | 10.4 ± 6.3 | 99.8 ± 0.3 | 13.4 ± 8.1 |
| 1200 | 100.0 ± 0.0 | 10.5 ± 6.6 | 99.2 ± 1.2 | 14.8 ± 7.7 |