T900
done · peak unseen_mismatch 99.9
α schedule
α 5.0 ×19 α 5.0 ×19 α 5.0 ×19 α 5.0 ×19 α 5.0 ×19 α 5.0 ×19 α 5.0 ×19 α 5.0 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 3.1429 ×19 α 3.1429 ×19 α 3.1429 ×19 α 3.1429 ×19 α 3.1429 ×19 α 3.1429 ×19 α 3.1429 ×19 α 3.1429 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×19 α 0.7 ×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 0.7 ×18 α 0.7 ×18 α 0.7 ×18 α 0.7 ×18 α 0.7 ×18 α 0.7 ×18 α 0.7 ×18 α 0.7 ×18
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.7
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 | 51.5 ± 1.5 | 49.9 ± 1.3 | 49.8 ± 1.9 | 49.6 ± 2.2 |
| 50 | 50.9 ± 1.8 | 50.0 ± 1.2 | 50.2 ± 1.6 | 50.4 ± 1.5 |
| 75 | 54.0 ± 1.9 | 51.6 ± 1.9 | 51.1 ± 1.7 | 50.1 ± 2.2 |
| 100 | 62.4 ± 6.2 | 58.5 ± 4.6 | 54.2 ± 4.3 | 51.8 ± 2.1 |
| 125 | 74.4 ± 8.7 | 70.9 ± 8.4 | 62.6 ± 8.9 | 59.0 ± 7.4 |
| 150 | 78.4 ± 4.5 | 76.4 ± 4.0 | 64.7 ± 6.3 | 61.9 ± 5.5 |
| 175 | 92.4 ± 6.1 | 89.1 ± 7.3 | 86.2 ± 10.1 | 81.4 ± 10.6 |
| 200 | 99.8 ± 0.4 | 96.6 ± 9.8 | 99.5 ± 0.8 | 95.5 ± 10.4 |
| 225 | 99.8 ± 0.9 | 93.5 ± 13.8 | 99.6 ± 1.6 | 92.8 ± 14.0 |
| 250 | 100.0 ± 0.0 | 93.8 ± 16.1 | 100.0 ± 0.2 | 93.2 ± 17.2 |
| 275 | 100.0 ± 0.0 | 94.8 ± 13.1 | 100.0 ± 0.0 | 94.0 ± 13.1 |
| 300 | 100.0 ± 0.0 | 96.9 ± 9.4 | 100.0 ± 0.0 | 96.4 ± 9.5 |
| 325 | 97.6 ± 1.6 | 94.9 ± 3.2 | 92.5 ± 4.6 | 85.8 ± 8.2 |
| 350 | 96.6 ± 1.6 | 93.3 ± 3.1 | 91.0 ± 4.4 | 84.0 ± 7.1 |
| 375 | 95.2 ± 1.2 | 92.4 ± 2.2 | 88.0 ± 3.9 | 81.5 ± 5.2 |
| 400 | 94.2 ± 1.5 | 91.3 ± 1.8 | 84.7 ± 2.9 | 78.2 ± 3.7 |
| 450 | 93.5 ± 1.4 | 91.2 ± 1.6 | 83.3 ± 2.4 | 77.1 ± 3.8 |
| 500 | 100.0 ± 0.0 | 100.0 ± 0.0 | 100.0 ± 0.2 | 99.9 ± 0.3 |
| 550 | 100.0 ± 0.0 | 99.9 ± 0.1 | 100.0 ± 0.0 | 99.6 ± 0.6 |
| 600 | 100.0 ± 0.0 | 99.4 ± 1.3 | 99.9 ± 0.3 | 97.3 ± 3.9 |
| 650 | 99.8 ± 0.2 | 98.6 ± 1.1 | 97.7 ± 1.4 | 91.3 ± 6.5 |
| 700 | 99.8 ± 0.3 | 98.8 ± 0.9 | 97.7 ± 1.3 | 91.3 ± 5.6 |
| 750 | 99.8 ± 0.1 | 99.2 ± 0.6 | 97.9 ± 1.4 | 92.5 ± 4.1 |
| 800 | 100.0 ± 0.0 | 99.7 ± 0.5 | 100.0 ± 0.1 | 96.0 ± 4.4 |
| 850 | 100.0 ± 0.0 | 99.2 ± 1.3 | 100.0 ± 0.1 | 93.8 ± 5.3 |
| 900 | 100.0 ± 0.0 | 97.4 ± 6.5 | 99.7 ± 0.6 | 89.8 ± 10.6 |