s3_e1
done · peak unseen_mismatch 94.9
α schedule
α 3.0 ×50 α 2.913 ×50 α 2.8261 ×50 α 2.7391 ×50 α 2.6522 ×50 α 2.5652 ×50 α 2.4783 ×50 α 2.3913 ×50 α 2.3043 ×50 α 2.2174 ×50 α 2.1304 ×50 α 2.0435 ×50 α 1.9565 ×50 α 1.8696 ×50 α 1.7826 ×50 α 1.6957 ×50 α 1.6087 ×50 α 1.5217 ×50 α 1.4348 ×50 α 1.3478 ×50 α 1.2609 ×50 α 1.1739 ×50 α 1.087 ×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 | 51.3 ± 2.2 | 49.6 ± 0.8 | 50.0 ± 1.7 | 50.3 ± 1.6 |
| 100 | 56.3 ± 6.5 | 53.6 ± 4.1 | 53.5 ± 5.5 | 50.5 ± 2.2 |
| 150 | 77.0 ± 15.2 | 74.6 ± 13.3 | 70.2 ± 15.4 | 65.6 ± 12.7 |
| 200 | 90.4 ± 4.2 | 88.3 ± 4.5 | 80.4 ± 7.0 | 76.4 ± 7.1 |
| 250 | 92.3 ± 2.2 | 91.3 ± 1.4 | 81.2 ± 5.0 | 76.7 ± 5.3 |
| 300 | 93.9 ± 1.4 | 92.2 ± 1.8 | 81.8 ± 4.2 | 77.0 ± 5.5 |
| 350 | 94.1 ± 1.6 | 92.1 ± 2.0 | 81.5 ± 4.0 | 76.3 ± 4.7 |
| 400 | 94.5 ± 1.5 | 92.3 ± 1.6 | 81.2 ± 4.8 | 76.0 ± 4.8 |
| 450 | 96.2 ± 1.4 | 93.8 ± 1.6 | 84.7 ± 5.2 | 77.7 ± 6.6 |
| 500 | 96.7 ± 1.1 | 94.9 ± 1.0 | 84.9 ± 3.3 | 78.6 ± 3.3 |
| 550 | 97.1 ± 1.3 | 95.2 ± 1.4 | 86.8 ± 5.4 | 80.6 ± 6.3 |
| 600 | 97.4 ± 1.2 | 95.3 ± 1.4 | 88.7 ± 5.4 | 81.0 ± 5.3 |
| 650 | 98.0 ± 0.9 | 96.2 ± 1.1 | 89.8 ± 3.9 | 81.9 ± 4.9 |
| 700 | 98.5 ± 0.7 | 96.6 ± 2.0 | 91.8 ± 2.7 | 83.8 ± 5.6 |
| 750 | 99.0 ± 0.4 | 97.4 ± 1.1 | 93.1 ± 2.5 | 85.2 ± 5.6 |
| 800 | 99.0 ± 0.4 | 97.4 ± 1.1 | 93.2 ± 3.5 | 85.1 ± 5.1 |
| 850 | 99.3 ± 0.3 | 97.7 ± 1.2 | 94.4 ± 2.3 | 85.6 ± 6.3 |
| 900 | 99.7 ± 0.3 | 98.8 ± 0.7 | 96.7 ± 1.5 | 89.2 ± 5.2 |
| 950 | 99.7 ± 0.3 | 98.4 ± 1.7 | 96.8 ± 2.6 | 88.1 ± 8.5 |
| 1000 | 99.9 ± 0.2 | 99.0 ± 0.9 | 97.8 ± 1.3 | 90.6 ± 3.6 |
| 1050 | 99.9 ± 0.1 | 99.4 ± 0.4 | 98.3 ± 0.7 | 92.5 ± 3.9 |
| 1100 | 99.9 ± 0.1 | 99.3 ± 0.9 | 98.5 ± 1.1 | 91.5 ± 5.6 |
| 1150 | 100.0 ± 0.0 | 99.6 ± 0.6 | 99.2 ± 0.5 | 94.9 ± 4.4 |
| 1200 | 100.0 ± 0.0 | 99.7 ± 0.4 | 99.5 ± 0.3 | 94.1 ± 4.5 |