peak_50
done · peak unseen_mismatch 97.3
α schedule
α 5.0 ×150 α 1.0 ×150 α 3.5714 ×150 α 1.0 ×150 α 2.1429 ×150 α 1.0 ×150
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 | 52.0 ± 1.2 | 50.6 ± 1.4 | 49.4 ± 1.7 | 49.7 ± 1.3 |
| 100 | 60.7 ± 8.9 | 53.6 ± 3.9 | 56.1 ± 8.8 | 50.4 ± 2.1 |
| 150 | 89.7 ± 8.3 | 84.7 ± 6.7 | 83.5 ± 12.6 | 76.3 ± 11.0 |
| 200 | 99.8 ± 0.4 | 89.5 ± 10.5 | 99.7 ± 0.8 | 88.1 ± 10.2 |
| 250 | 100.0 ± 0.1 | 87.9 ± 20.6 | 100.0 ± 0.0 | 86.8 ± 20.7 |
| 300 | 100.0 ± 0.1 | 92.5 ± 14.3 | 99.9 ± 0.3 | 90.4 ± 15.6 |
| 350 | 98.1 ± 1.9 | 94.3 ± 4.7 | 95.8 ± 3.9 | 87.7 ± 9.1 |
| 400 | 96.1 ± 2.8 | 93.1 ± 3.5 | 90.9 ± 6.1 | 84.0 ± 9.6 |
| 450 | 94.6 ± 2.8 | 91.5 ± 3.5 | 87.3 ± 6.9 | 78.8 ± 8.6 |
| 500 | 100.0 ± 0.0 | 99.4 ± 1.2 | 99.8 ± 0.5 | 97.3 ± 4.9 |
| 550 | 100.0 ± 0.1 | 99.2 ± 0.7 | 99.8 ± 0.7 | 95.0 ± 3.0 |
| 600 | 100.0 ± 0.1 | 99.2 ± 1.2 | 99.8 ± 0.4 | 95.1 ± 6.2 |
| 650 | 98.0 ± 1.0 | 94.7 ± 2.1 | 91.9 ± 4.4 | 82.6 ± 7.6 |
| 700 | 97.5 ± 0.7 | 94.1 ± 2.3 | 91.6 ± 3.7 | 80.7 ± 8.6 |
| 750 | 97.4 ± 1.1 | 93.8 ± 2.0 | 90.7 ± 3.8 | 79.3 ± 7.0 |
| 800 | 100.0 ± 0.1 | 98.7 ± 1.4 | 99.5 ± 0.4 | 91.1 ± 7.7 |
| 850 | 100.0 ± 0.0 | 98.1 ± 2.9 | 99.6 ± 0.4 | 91.3 ± 8.1 |
| 900 | 100.0 ± 0.0 | 97.7 ± 2.6 | 99.4 ± 0.9 | 87.7 ± 10.2 |