fixed_16
done · peak unseen_mismatch 81.9
Referenced by
α schedule
α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25 α 1.6 ×25
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 | 50.9 ± 1.5 | 49.9 ± 1.6 | 50.2 ± 1.8 | 49.9 ± 1.6 |
| 50 | 50.8 ± 1.7 | 49.7 ± 1.7 | 50.2 ± 1.5 | 49.9 ± 1.5 |
| 75 | 50.7 ± 1.7 | 50.0 ± 1.7 | 50.2 ± 1.8 | 49.9 ± 1.5 |
| 100 | 51.6 ± 2.7 | 49.6 ± 1.6 | 50.6 ± 2.4 | 49.6 ± 1.8 |
| 125 | 54.3 ± 7.5 | 49.5 ± 3.4 | 53.3 ± 7.6 | 49.4 ± 3.1 |
| 150 | 61.6 ± 15.4 | 48.7 ± 7.3 | 60.3 ± 15.6 | 48.4 ± 7.0 |
| 175 | 71.4 ± 20.1 | 50.2 ± 10.8 | 70.2 ± 20.8 | 49.3 ± 10.5 |
| 200 | 81.8 ± 19.8 | 52.1 ± 16.9 | 80.9 ± 20.6 | 51.1 ± 16.3 |
| 225 | 91.3 ± 15.2 | 58.3 ± 22.2 | 90.6 ± 16.1 | 57.5 ± 21.4 |
| 250 | 97.1 ± 9.6 | 60.7 ± 24.8 | 96.7 ± 10.5 | 60.2 ± 24.3 |
| 275 | 98.4 ± 8.0 | 69.3 ± 26.4 | 98.3 ± 8.4 | 68.4 ± 25.9 |
| 300 | 99.0 ± 5.8 | 73.8 ± 25.2 | 98.9 ± 6.2 | 72.7 ± 25.0 |
| 325 | 99.4 ± 4.4 | 78.2 ± 25.0 | 99.3 ± 4.4 | 77.3 ± 24.6 |
| 350 | 99.6 ± 2.7 | 82.9 ± 23.2 | 99.5 ± 3.2 | 81.9 ± 22.9 |