s3_e2
done · peak unseen_mismatch 86.1
α schedule
α 3.0 ×50 α 2.9565 ×50 α 2.913 ×50 α 2.8696 ×50 α 2.8261 ×50 α 2.7826 ×50 α 2.7391 ×50 α 2.6957 ×50 α 2.6522 ×50 α 2.6087 ×50 α 2.5652 ×50 α 2.5217 ×50 α 2.4783 ×50 α 2.4348 ×50 α 2.3913 ×50 α 2.3478 ×50 α 2.3043 ×50 α 2.2609 ×50 α 2.2174 ×50 α 2.1739 ×50 α 2.1304 ×50 α 2.087 ×50 α 2.0435 ×50 α 2.0 ×50
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 2.0
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 | 56.4 ± 6.5 | 53.6 ± 4.4 | 53.8 ± 5.9 | 50.6 ± 2.5 |
| 150 | 76.6 ± 14.9 | 74.7 ± 13.5 | 69.3 ± 13.9 | 64.6 ± 12.3 |
| 200 | 89.9 ± 3.5 | 87.8 ± 3.8 | 78.7 ± 6.3 | 75.0 ± 6.2 |
| 250 | 91.1 ± 1.9 | 89.7 ± 1.2 | 77.8 ± 3.8 | 74.9 ± 3.3 |
| 300 | 92.5 ± 1.6 | 91.2 ± 0.9 | 77.6 ± 3.5 | 74.5 ± 3.6 |
| 350 | 92.9 ± 2.3 | 91.3 ± 1.7 | 78.7 ± 5.0 | 74.7 ± 4.6 |
| 400 | 92.8 ± 1.7 | 90.8 ± 1.5 | 77.9 ± 4.9 | 72.5 ± 4.2 |
| 450 | 94.3 ± 1.3 | 92.3 ± 1.3 | 80.6 ± 6.2 | 74.4 ± 6.4 |
| 500 | 94.8 ± 1.2 | 92.8 ± 1.5 | 81.0 ± 4.3 | 75.3 ± 4.1 |
| 550 | 94.8 ± 1.6 | 93.3 ± 0.9 | 82.3 ± 4.9 | 76.9 ± 5.9 |
| 600 | 95.3 ± 1.5 | 94.0 ± 1.7 | 83.2 ± 5.4 | 77.6 ± 5.9 |
| 650 | 95.9 ± 1.6 | 94.5 ± 0.8 | 83.0 ± 6.8 | 78.5 ± 6.7 |
| 700 | 95.9 ± 1.1 | 94.4 ± 1.5 | 84.4 ± 1.6 | 78.5 ± 2.5 |
| 750 | 96.7 ± 0.8 | 94.9 ± 1.0 | 86.3 ± 3.9 | 81.1 ± 3.4 |
| 800 | 96.8 ± 0.7 | 95.1 ± 0.8 | 86.6 ± 3.2 | 80.6 ± 4.6 |
| 850 | 97.0 ± 0.6 | 95.6 ± 0.9 | 86.9 ± 2.9 | 81.1 ± 4.1 |
| 900 | 97.7 ± 1.0 | 96.2 ± 0.7 | 87.5 ± 4.0 | 82.0 ± 4.4 |
| 950 | 97.3 ± 1.1 | 96.0 ± 0.9 | 87.4 ± 2.5 | 82.7 ± 3.0 |
| 1000 | 98.0 ± 0.5 | 96.7 ± 0.5 | 89.6 ± 2.6 | 83.8 ± 3.3 |
| 1050 | 97.9 ± 0.6 | 96.9 ± 0.9 | 89.4 ± 3.6 | 84.0 ± 5.2 |
| 1100 | 98.0 ± 0.6 | 96.6 ± 0.6 | 90.5 ± 2.0 | 84.8 ± 2.0 |
| 1150 | 98.6 ± 0.6 | 97.7 ± 0.7 | 91.2 ± 2.3 | 86.1 ± 2.5 |
| 1200 | 98.6 ± 0.4 | 97.5 ± 0.8 | 90.9 ± 2.7 | 84.7 ± 5.7 |