smooth_K6
done · peak unseen_mismatch 85.1
α schedule
α 2.8571 ×150 α 2.5714 ×150 α 2.2857 ×150 α 2.0 ×150 α 1.7143 ×150 α 1.4286 ×150
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.4286
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 ± 1.6 | 50.2 ± 1.7 | 49.5 ± 1.4 | 50.1 ± 1.5 |
| 50 | 51.2 ± 1.8 | 49.9 ± 1.8 | 49.9 ± 1.8 | 49.4 ± 1.3 |
| 75 | 52.1 ± 2.0 | 49.8 ± 2.0 | 51.3 ± 2.0 | 50.0 ± 1.7 |
| 100 | 55.6 ± 8.0 | 49.1 ± 4.1 | 54.1 ± 7.2 | 49.1 ± 4.6 |
| 125 | 66.8 ± 16.2 | 51.6 ± 7.9 | 64.1 ± 16.2 | 50.0 ± 6.1 |
| 150 | 83.8 ± 15.2 | 66.0 ± 14.8 | 81.0 ± 16.3 | 62.8 ± 14.1 |
| 175 | 96.8 ± 4.0 | 78.6 ± 21.6 | 95.3 ± 4.9 | 76.1 ± 19.9 |
| 200 | 99.0 ± 1.2 | 86.1 ± 19.8 | 98.0 ± 2.8 | 83.4 ± 18.8 |
| 225 | 98.5 ± 1.6 | 88.7 ± 16.8 | 97.0 ± 3.4 | 84.7 ± 16.3 |
| 250 | 98.2 ± 1.7 | 88.8 ± 16.9 | 95.6 ± 4.2 | 84.1 ± 15.8 |
| 275 | 97.7 ± 1.8 | 89.5 ± 15.1 | 93.8 ± 5.3 | 82.6 ± 15.2 |
| 300 | 97.1 ± 2.3 | 91.0 ± 10.7 | 92.4 ± 5.3 | 83.6 ± 10.5 |
| 325 | 97.0 ± 2.1 | 91.6 ± 7.3 | 91.2 ± 6.8 | 83.0 ± 9.0 |
| 350 | 97.1 ± 1.9 | 94.5 ± 3.3 | 91.3 ± 5.9 | 85.1 ± 8.8 |
| 375 | 96.9 ± 2.0 | 93.4 ± 3.2 | 90.1 ± 6.3 | 83.1 ± 8.4 |
| 400 | 96.7 ± 1.8 | 94.3 ± 2.8 | 89.5 ± 6.0 | 82.8 ± 8.7 |
| 450 | 96.3 ± 1.6 | 93.0 ± 2.5 | 88.1 ± 5.7 | 79.8 ± 8.0 |
| 500 | 97.6 ± 1.3 | 94.7 ± 2.5 | 90.7 ± 4.7 | 82.2 ± 8.1 |
| 550 | 97.5 ± 1.2 | 94.3 ± 2.7 | 90.3 ± 3.9 | 79.3 ± 9.0 |
| 600 | 97.5 ± 0.8 | 94.5 ± 2.5 | 89.5 ± 4.1 | 78.5 ± 8.8 |
| 650 | 98.7 ± 0.8 | 95.6 ± 2.7 | 94.1 ± 3.2 | 81.9 ± 10.0 |
| 700 | 98.8 ± 0.5 | 95.9 ± 2.3 | 93.6 ± 3.2 | 81.8 ± 10.3 |
| 750 | 98.9 ± 0.6 | 95.4 ± 2.0 | 93.5 ± 2.7 | 78.8 ± 8.1 |
| 800 | 99.6 ± 0.3 | 97.2 ± 1.7 | 96.7 ± 1.4 | 82.9 ± 9.9 |
| 850 | 99.8 ± 0.2 | 96.7 ± 3.1 | 97.7 ± 1.5 | 82.0 ± 12.2 |
| 900 | 99.7 ± 0.4 | 96.3 ± 2.6 | 96.7 ± 2.0 | 80.0 ± 10.8 |