s0_e1
done · peak unseen_mismatch 49.7
α schedule
α 0.0 ×50 α 0.0435 ×50 α 0.087 ×50 α 0.1304 ×50 α 0.1739 ×50 α 0.2174 ×50 α 0.2609 ×50 α 0.3043 ×50 α 0.3478 ×50 α 0.3913 ×50 α 0.4348 ×50 α 0.4783 ×50 α 0.5217 ×50 α 0.5652 ×50 α 0.6087 ×50 α 0.6522 ×50 α 0.6957 ×50 α 0.7391 ×50 α 0.7826 ×50 α 0.8261 ×50 α 0.8696 ×50 α 0.913 ×50 α 0.9565 ×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 | 50.4 ± 1.6 | 49.9 ± 1.1 | 49.8 ± 1.9 | 49.6 ± 1.9 |
| 100 | 50.2 ± 1.4 | 50.0 ± 2.0 | 49.6 ± 1.1 | 49.7 ± 1.1 |
| 150 | 49.4 ± 2.2 | 49.5 ± 1.8 | 50.2 ± 1.4 | 48.5 ± 1.6 |
| 200 | 57.6 ± 10.4 | 45.8 ± 7.0 | 56.7 ± 11.4 | 45.5 ± 6.9 |
| 250 | 65.4 ± 22.5 | 47.1 ± 10.9 | 65.3 ± 22.4 | 46.8 ± 10.7 |
| 300 | 72.0 ± 23.1 | 37.5 ± 16.9 | 71.4 ± 23.4 | 38.7 ± 17.9 |
| 350 | 81.7 ± 22.5 | 35.5 ± 17.4 | 80.9 ± 23.5 | 35.9 ± 17.4 |
| 400 | 98.2 ± 5.3 | 19.3 ± 17.3 | 98.0 ± 6.1 | 20.1 ± 17.7 |
| 450 | 100.0 ± 0.0 | 32.2 ± 24.1 | 100.0 ± 0.1 | 32.0 ± 23.4 |
| 500 | 100.0 ± 0.0 | 19.5 ± 18.9 | 100.0 ± 0.0 | 19.9 ± 19.8 |
| 550 | 100.0 ± 0.0 | 32.1 ± 35.2 | 100.0 ± 0.0 | 32.2 ± 35.4 |
| 600 | 100.0 ± 0.0 | 12.7 ± 18.3 | 100.0 ± 0.0 | 13.3 ± 19.0 |
| 650 | 100.0 ± 0.0 | 17.5 ± 20.6 | 100.0 ± 0.0 | 17.8 ± 21.2 |
| 700 | 100.0 ± 0.0 | 23.0 ± 25.1 | 100.0 ± 0.0 | 23.6 ± 26.2 |
| 750 | 100.0 ± 0.0 | 22.2 ± 21.7 | 100.0 ± 0.0 | 22.3 ± 21.6 |
| 800 | 100.0 ± 0.0 | 20.6 ± 23.5 | 100.0 ± 0.0 | 21.6 ± 24.3 |
| 850 | 100.0 ± 0.0 | 21.5 ± 23.4 | 100.0 ± 0.0 | 21.9 ± 23.3 |
| 900 | 100.0 ± 0.0 | 25.7 ± 25.0 | 100.0 ± 0.0 | 25.6 ± 23.3 |
| 950 | 100.0 ± 0.0 | 39.1 ± 29.5 | 100.0 ± 0.0 | 39.2 ± 29.5 |
| 1000 | 100.0 ± 0.0 | 32.4 ± 31.3 | 100.0 ± 0.0 | 32.8 ± 31.7 |
| 1050 | 100.0 ± 0.0 | 35.4 ± 30.0 | 100.0 ± 0.0 | 36.0 ± 30.1 |
| 1100 | 100.0 ± 0.0 | 27.9 ± 23.3 | 100.0 ± 0.0 | 28.2 ± 22.9 |
| 1150 | 100.0 ± 0.0 | 33.8 ± 27.6 | 100.0 ± 0.0 | 32.6 ± 26.8 |
| 1200 | 100.0 ± 0.0 | 21.3 ± 20.7 | 100.0 ± 0.0 | 21.4 ± 20.9 |