duty_0.75
done · peak unseen_mismatch 98.5
α schedule
α 3.0 ×19 α 3.0 ×19 α 3.0 ×19 α 3.0 ×19 α 3.0 ×19 α 3.0 ×19 α 3.0 ×19 α 3.0 ×19 α 3.0 ×19 α 3.0 ×19 α 3.0 ×19 α 3.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 1.0 ×19 α 2.1429 ×19 α 2.1429 ×19 α 2.1429 ×19 α 2.1429 ×19 α 2.1429 ×19 α 2.1429 ×19 α 2.1429 ×19 α 2.1429 ×19 α 2.1429 ×19 α 2.1429 ×19 α 2.1429 ×19 α 2.1429 ×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 1.2857 ×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
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 | 51.4 ± 1.5 | 49.5 ± 1.6 | 49.8 ± 1.9 | 49.7 ± 2.1 |
| 50 | 50.7 ± 1.9 | 49.6 ± 1.4 | 50.1 ± 1.5 | 50.8 ± 1.1 |
| 75 | 51.9 ± 1.8 | 50.1 ± 1.9 | 50.5 ± 1.7 | 50.0 ± 1.9 |
| 100 | 53.8 ± 2.9 | 52.1 ± 2.9 | 51.5 ± 1.8 | 50.0 ± 1.5 |
| 125 | 65.8 ± 10.5 | 60.5 ± 8.1 | 59.9 ± 9.2 | 54.9 ± 5.3 |
| 150 | 79.3 ± 12.4 | 73.6 ± 12.8 | 72.2 ± 14.1 | 66.3 ± 12.7 |
| 175 | 86.1 ± 10.6 | 82.1 ± 10.4 | 77.2 ± 13.5 | 72.0 ± 11.7 |
| 200 | 88.3 ± 9.6 | 85.7 ± 9.2 | 78.2 ± 11.1 | 74.4 ± 9.9 |
| 225 | 88.8 ± 9.4 | 87.2 ± 8.9 | 77.4 ± 11.4 | 74.3 ± 11.1 |
| 250 | 96.7 ± 9.9 | 95.9 ± 9.7 | 95.1 ± 9.8 | 93.1 ± 10.2 |
| 275 | 98.5 ± 5.3 | 97.7 ± 7.1 | 97.6 ± 5.5 | 95.3 ± 6.7 |
| 300 | 99.9 ± 0.2 | 96.1 ± 14.1 | 99.0 ± 1.0 | 93.9 ± 14.2 |
| 325 | 96.4 ± 1.8 | 92.2 ± 7.5 | 87.5 ± 6.2 | 80.0 ± 7.3 |
| 350 | 97.1 ± 1.3 | 94.3 ± 2.1 | 89.4 ± 4.7 | 82.8 ± 6.2 |
| 375 | 96.8 ± 1.1 | 94.5 ± 2.1 | 89.1 ± 3.7 | 82.5 ± 5.9 |
| 400 | 96.9 ± 1.1 | 95.0 ± 1.8 | 88.5 ± 3.7 | 82.9 ± 5.2 |
| 450 | 96.9 ± 0.9 | 95.0 ± 1.6 | 89.1 ± 3.6 | 83.1 ± 5.3 |
| 500 | 97.3 ± 0.9 | 95.1 ± 1.6 | 89.7 ± 4.1 | 83.3 ± 5.0 |
| 550 | 100.0 ± 0.1 | 99.8 ± 0.2 | 99.7 ± 0.5 | 98.5 ± 1.9 |
| 600 | 100.0 ± 0.1 | 99.5 ± 0.5 | 99.1 ± 1.0 | 95.1 ± 5.0 |
| 650 | 99.8 ± 0.2 | 99.0 ± 0.7 | 97.5 ± 1.6 | 91.5 ± 5.9 |
| 700 | 99.8 ± 0.1 | 98.9 ± 0.7 | 98.0 ± 1.1 | 89.9 ± 6.4 |
| 750 | 99.8 ± 0.2 | 98.9 ± 1.0 | 97.8 ± 1.4 | 90.2 ± 7.0 |
| 800 | 99.8 ± 0.2 | 98.7 ± 1.0 | 97.5 ± 1.3 | 89.0 ± 6.6 |
| 850 | 100.0 ± 0.0 | 99.4 ± 0.8 | 99.8 ± 0.4 | 94.6 ± 6.5 |
| 900 | 100.0 ± 0.0 | 99.2 ± 0.9 | 99.2 ± 0.9 | 90.9 ± 7.7 |