peak_10
done · peak unseen_mismatch 50.9
α schedule
α 1.0 ×150 α 1.0 ×150 α 0.7143 ×150 α 1.0 ×150 α 0.4286 ×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 | 50.5 ± 1.2 | 49.9 ± 1.3 | 50.0 ± 1.6 | 49.2 ± 1.3 |
| 100 | 50.3 ± 1.2 | 48.9 ± 1.2 | 50.3 ± 1.7 | 50.9 ± 2.2 |
| 150 | 51.4 ± 1.6 | 50.1 ± 1.3 | 50.4 ± 2.0 | 50.4 ± 2.1 |
| 200 | 57.8 ± 14.2 | 46.3 ± 7.2 | 56.8 ± 14.6 | 45.5 ± 7.2 |
| 250 | 74.8 ± 23.8 | 35.9 ± 20.5 | 74.5 ± 23.7 | 35.4 ± 19.6 |
| 300 | 75.8 ± 24.3 | 30.8 ± 22.8 | 75.7 ± 24.5 | 29.7 ± 22.1 |
| 350 | 85.6 ± 21.8 | 26.3 ± 20.9 | 84.8 ± 23.1 | 25.8 ± 20.5 |
| 400 | 90.8 ± 18.5 | 18.4 ± 20.7 | 90.9 ± 18.3 | 18.2 ± 20.6 |
| 450 | 96.1 ± 11.8 | 15.0 ± 18.2 | 96.1 ± 11.8 | 14.5 ± 18.1 |
| 500 | 100.0 ± 0.0 | 14.3 ± 17.1 | 100.0 ± 0.0 | 14.1 ± 17.1 |
| 550 | 100.0 ± 0.0 | 21.9 ± 19.1 | 100.0 ± 0.0 | 21.7 ± 19.1 |
| 600 | 100.0 ± 0.0 | 24.0 ± 22.7 | 100.0 ± 0.0 | 24.1 ± 22.9 |
| 650 | 100.0 ± 0.0 | 20.5 ± 21.1 | 100.0 ± 0.0 | 21.4 ± 21.4 |
| 700 | 100.0 ± 0.0 | 21.2 ± 16.5 | 100.0 ± 0.0 | 20.8 ± 17.1 |
| 750 | 100.0 ± 0.0 | 25.7 ± 22.9 | 100.0 ± 0.0 | 24.8 ± 22.9 |
| 800 | 100.0 ± 0.0 | 32.9 ± 29.8 | 100.0 ± 0.0 | 32.5 ± 29.9 |
| 850 | 100.0 ± 0.0 | 29.2 ± 25.7 | 100.0 ± 0.0 | 29.0 ± 25.4 |
| 900 | 100.0 ± 0.0 | 32.3 ± 28.2 | 100.0 ± 0.0 | 32.4 ± 28.3 |