fixed_16
done · peak unseen_mismatch 91.5
Referenced by
α schedule
α 1.6 ×900
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.6
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 |
|---|---|---|---|---|
| 25 | 50.7 ± 1.6 | 49.4 ± 1.6 | 49.9 ± 1.9 | 50.5 ± 1.4 |
| 50 | 51.2 ± 1.7 | 49.7 ± 1.8 | 49.7 ± 1.6 | 49.6 ± 1.7 |
| 75 | 51.4 ± 1.9 | 49.9 ± 1.3 | 50.6 ± 1.8 | 49.7 ± 1.4 |
| 100 | 52.6 ± 3.0 | 49.6 ± 2.0 | 51.3 ± 2.7 | 49.5 ± 1.9 |
| 125 | 56.7 ± 11.1 | 47.5 ± 6.3 | 55.6 ± 11.2 | 47.4 ± 6.6 |
| 150 | 64.9 ± 18.1 | 44.6 ± 12.0 | 63.9 ± 18.4 | 44.9 ± 11.7 |
| 175 | 73.5 ± 21.5 | 45.3 ± 14.1 | 73.2 ± 21.8 | 44.7 ± 13.8 |
| 200 | 84.7 ± 20.2 | 44.2 ± 17.9 | 84.4 ± 20.6 | 43.9 ± 17.6 |
| 225 | 90.7 ± 17.5 | 37.3 ± 22.5 | 90.4 ± 17.7 | 37.4 ± 22.1 |
| 250 | 94.6 ± 14.3 | 33.1 ± 24.2 | 94.3 ± 15.1 | 32.9 ± 23.6 |
| 275 | 97.4 ± 9.0 | 38.9 ± 25.0 | 97.1 ± 9.7 | 38.8 ± 24.9 |
| 300 | 99.3 ± 3.3 | 37.4 ± 32.0 | 99.3 ± 3.4 | 37.9 ± 31.5 |
| 325 | 100.0 ± 0.0 | 41.9 ± 34.2 | 100.0 ± 0.0 | 41.9 ± 33.9 |
| 350 | 100.0 ± 0.0 | 46.6 ± 31.9 | 100.0 ± 0.0 | 46.3 ± 31.6 |
| 375 | 100.0 ± 0.0 | 50.6 ± 35.6 | 100.0 ± 0.0 | 50.4 ± 35.0 |
| 400 | 100.0 ± 0.0 | 57.9 ± 35.8 | 99.9 ± 0.3 | 57.3 ± 35.1 |
| 450 | 100.0 ± 0.1 | 64.9 ± 35.9 | 99.9 ± 0.3 | 64.7 ± 35.3 |
| 500 | 100.0 ± 0.0 | 74.6 ± 33.1 | 99.9 ± 0.3 | 74.3 ± 32.7 |
| 550 | 99.9 ± 0.3 | 79.2 ± 28.5 | 99.9 ± 0.4 | 78.3 ± 28.2 |
| 600 | 99.9 ± 0.1 | 84.2 ± 25.9 | 99.7 ± 1.1 | 83.2 ± 25.2 |
| 650 | 99.9 ± 0.3 | 90.0 ± 22.8 | 99.5 ± 1.0 | 88.0 ± 22.7 |
| 700 | 99.8 ± 0.3 | 92.2 ± 16.7 | 98.9 ± 1.7 | 89.4 ± 16.5 |
| 750 | 99.7 ± 0.5 | 93.9 ± 12.9 | 98.9 ± 1.7 | 91.2 ± 13.0 |
| 800 | 99.7 ± 0.4 | 94.6 ± 11.3 | 98.6 ± 1.6 | 91.5 ± 11.9 |
| 850 | 99.5 ± 0.5 | 94.0 ± 16.1 | 98.0 ± 2.3 | 90.1 ± 16.0 |
| 900 | 99.4 ± 0.6 | 93.4 ± 17.1 | 97.9 ± 2.1 | 88.8 ± 17.1 |