peak_15
done · peak unseen_mismatch 49.6
α schedule
α 1.5 ×150 α 1.0 ×150 α 1.0714 ×150 α 1.0 ×150 α 0.6429 ×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 | 50.9 ± 1.9 | 49.5 ± 1.6 | 49.2 ± 1.6 | 49.6 ± 2.3 |
| 100 | 51.3 ± 0.7 | 49.8 ± 1.4 | 50.3 ± 1.0 | 48.9 ± 1.4 |
| 150 | 52.8 ± 3.6 | 48.7 ± 1.6 | 51.9 ± 2.9 | 49.1 ± 1.9 |
| 200 | 67.8 ± 21.5 | 36.6 ± 17.4 | 67.7 ± 21.9 | 36.4 ± 17.1 |
| 250 | 82.2 ± 22.7 | 27.1 ± 20.6 | 83.2 ± 21.4 | 27.4 ± 19.6 |
| 300 | 86.5 ± 20.9 | 35.5 ± 22.8 | 86.4 ± 21.0 | 35.8 ± 22.7 |
| 350 | 97.2 ± 7.4 | 35.8 ± 26.2 | 97.0 ± 7.9 | 36.5 ± 25.7 |
| 400 | 100.0 ± 0.0 | 22.2 ± 18.8 | 100.0 ± 0.0 | 21.9 ± 18.6 |
| 450 | 100.0 ± 0.0 | 14.5 ± 17.6 | 100.0 ± 0.0 | 14.9 ± 17.4 |
| 500 | 100.0 ± 0.0 | 26.4 ± 27.1 | 100.0 ± 0.0 | 26.7 ± 27.5 |
| 550 | 100.0 ± 0.0 | 26.9 ± 31.8 | 100.0 ± 0.0 | 27.0 ± 32.6 |
| 600 | 100.0 ± 0.0 | 19.6 ± 27.4 | 100.0 ± 0.0 | 19.9 ± 27.5 |
| 650 | 100.0 ± 0.0 | 33.9 ± 27.9 | 100.0 ± 0.0 | 34.5 ± 28.4 |
| 700 | 100.0 ± 0.0 | 23.3 ± 27.1 | 100.0 ± 0.0 | 23.3 ± 27.0 |
| 750 | 100.0 ± 0.0 | 32.4 ± 42.2 | 100.0 ± 0.0 | 32.8 ± 41.8 |
| 800 | 100.0 ± 0.0 | 43.0 ± 41.2 | 100.0 ± 0.0 | 42.7 ± 40.4 |
| 850 | 100.0 ± 0.0 | 41.6 ± 43.3 | 100.0 ± 0.0 | 41.4 ± 42.9 |
| 900 | 100.0 ± 0.0 | 40.9 ± 31.4 | 100.0 ± 0.0 | 41.1 ± 30.8 |