fixed_16
done · peak unseen_mismatch 86.6
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 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 | 50.7 ± 1.7 | 49.6 ± 1.6 | 50.0 ± 2.0 | 50.2 ± 1.8 |
| 50 | 50.9 ± 1.9 | 50.3 ± 1.4 | 50.4 ± 1.6 | 49.9 ± 1.7 |
| 75 | 51.1 ± 2.1 | 49.9 ± 1.8 | 50.2 ± 1.7 | 49.1 ± 1.5 |
| 100 | 51.6 ± 2.6 | 50.0 ± 1.5 | 50.8 ± 2.3 | 50.0 ± 1.3 |
| 125 | 55.4 ± 8.3 | 48.5 ± 3.6 | 54.5 ± 8.3 | 48.6 ± 3.6 |
| 150 | 63.3 ± 17.9 | 48.3 ± 10.9 | 63.0 ± 18.0 | 48.1 ± 10.0 |
| 175 | 70.3 ± 19.9 | 46.2 ± 12.2 | 69.5 ± 19.9 | 45.8 ± 11.7 |
| 200 | 79.6 ± 20.1 | 49.5 ± 16.4 | 79.1 ± 20.4 | 49.3 ± 16.1 |
| 225 | 93.8 ± 12.0 | 58.4 ± 22.5 | 93.1 ± 12.7 | 57.5 ± 21.3 |
| 250 | 97.7 ± 8.5 | 60.1 ± 27.2 | 97.5 ± 8.7 | 59.7 ± 26.5 |
| 275 | 98.8 ± 6.2 | 56.2 ± 28.8 | 98.4 ± 7.7 | 56.3 ± 28.4 |
| 300 | 99.9 ± 0.4 | 67.8 ± 26.1 | 99.9 ± 0.4 | 67.7 ± 25.9 |
| 325 | 100.0 ± 0.0 | 77.0 ± 24.0 | 99.9 ± 0.3 | 76.5 ± 23.6 |
| 350 | 100.0 ± 0.1 | 84.1 ± 19.1 | 99.8 ± 0.7 | 83.1 ± 18.9 |
| 375 | 99.9 ± 0.2 | 87.5 ± 17.2 | 99.7 ± 0.7 | 86.6 ± 17.2 |
| 400 | 99.9 ± 0.3 | 84.5 ± 20.0 | 99.6 ± 0.9 | 84.2 ± 19.2 |