start_2
done · peak unseen_mismatch 96.7
α schedule
α 2.0 ×19 α 2.0 ×19 α 2.0 ×19 α 2.0 ×19 α 2.0 ×19 α 2.0 ×19 α 2.0 ×19 α 2.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.6429 ×19 α 1.6429 ×19 α 1.6429 ×19 α 1.6429 ×19 α 1.6429 ×19 α 1.6429 ×19 α 1.6429 ×19 α 1.6429 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.2857 ×19 α 1.2857 ×19 α 1.2857 ×19 α 1.2857 ×19 α 1.2857 ×18 α 1.2857 ×18 α 1.2857 ×18 α 1.2857 ×18 α 1.0 ×18 α 1.0 ×18 α 1.0 ×18 α 1.0 ×18 α 1.0 ×18 α 1.0 ×18 α 1.0 ×18 α 1.0 ×18
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 |
|---|---|---|---|---|
| 25 | 51.6 ± 1.1 | 49.3 ± 1.6 | 49.7 ± 2.0 | 49.7 ± 2.2 |
| 50 | 50.6 ± 1.5 | 49.2 ± 1.1 | 50.1 ± 1.5 | 50.3 ± 1.2 |
| 75 | 51.5 ± 1.7 | 50.1 ± 1.8 | 50.4 ± 2.1 | 49.9 ± 1.8 |
| 100 | 51.3 ± 2.0 | 49.9 ± 1.4 | 50.5 ± 1.8 | 49.0 ± 1.1 |
| 125 | 54.7 ± 5.5 | 49.6 ± 3.2 | 53.3 ± 4.9 | 49.4 ± 2.6 |
| 150 | 67.4 ± 13.0 | 53.2 ± 7.1 | 64.9 ± 13.1 | 52.6 ± 6.1 |
| 175 | 83.1 ± 19.3 | 54.8 ± 19.3 | 82.0 ± 19.9 | 54.9 ± 19.0 |
| 200 | 92.2 ± 14.9 | 50.1 ± 21.8 | 91.6 ± 15.6 | 49.4 ± 21.4 |
| 225 | 98.1 ± 7.4 | 42.1 ± 22.3 | 98.0 ± 7.1 | 41.9 ± 22.1 |
| 250 | 100.0 ± 0.0 | 34.8 ± 28.5 | 100.0 ± 0.0 | 35.3 ± 28.0 |
| 275 | 100.0 ± 0.0 | 36.2 ± 28.3 | 100.0 ± 0.0 | 36.4 ± 27.9 |
| 300 | 100.0 ± 0.0 | 34.1 ± 30.1 | 100.0 ± 0.0 | 34.1 ± 30.2 |
| 325 | 100.0 ± 0.1 | 46.0 ± 32.4 | 100.0 ± 0.0 | 46.4 ± 32.5 |
| 350 | 100.0 ± 0.1 | 53.0 ± 28.4 | 100.0 ± 0.0 | 52.4 ± 28.1 |
| 375 | 100.0 ± 0.1 | 68.5 ± 30.7 | 99.8 ± 0.5 | 67.6 ± 29.9 |
| 400 | 99.9 ± 0.1 | 79.5 ± 29.4 | 99.9 ± 0.3 | 78.7 ± 29.4 |
| 450 | 99.9 ± 0.2 | 86.0 ± 23.3 | 99.5 ± 0.8 | 84.7 ± 23.0 |
| 500 | 100.0 ± 0.0 | 86.2 ± 23.9 | 99.9 ± 0.2 | 85.1 ± 23.5 |
| 550 | 100.0 ± 0.0 | 83.3 ± 23.4 | 99.9 ± 0.3 | 81.7 ± 23.2 |
| 600 | 100.0 ± 0.0 | 85.0 ± 24.3 | 99.9 ± 0.3 | 82.7 ± 24.5 |
| 650 | 99.9 ± 0.1 | 95.8 ± 6.4 | 99.3 ± 1.2 | 92.7 ± 7.0 |
| 700 | 99.9 ± 0.1 | 97.6 ± 3.3 | 99.3 ± 1.2 | 94.4 ± 4.8 |
| 750 | 99.9 ± 0.1 | 98.2 ± 3.6 | 98.9 ± 2.0 | 95.9 ± 5.6 |
| 800 | 100.0 ± 0.1 | 98.4 ± 2.7 | 99.5 ± 0.8 | 95.2 ± 7.0 |
| 850 | 100.0 ± 0.1 | 98.8 ± 2.3 | 99.4 ± 1.0 | 96.7 ± 3.9 |
| 900 | 100.0 ± 0.1 | 99.3 ± 1.0 | 99.5 ± 0.8 | 96.7 ± 3.7 |