s2_e5_blk
done · peak unseen_mismatch 96.5
α schedule
α 2.0 ×300 α 3.0 ×300 α 4.0 ×300 α 5.0 ×300
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 5.0
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 |
|---|---|---|---|---|
| 50 | 50.5 ± 1.9 | 49.9 ± 0.5 | 50.1 ± 1.9 | 50.1 ± 1.5 |
| 100 | 51.8 ± 1.7 | 50.8 ± 1.9 | 50.5 ± 1.7 | 50.4 ± 1.9 |
| 150 | 63.7 ± 13.6 | 55.5 ± 7.2 | 62.4 ± 12.6 | 52.6 ± 6.4 |
| 200 | 80.6 ± 21.2 | 67.9 ± 17.0 | 79.9 ± 21.9 | 66.8 ± 15.8 |
| 250 | 96.6 ± 6.5 | 91.5 ± 8.4 | 95.5 ± 8.6 | 88.7 ± 10.9 |
| 300 | 99.3 ± 0.7 | 98.5 ± 0.9 | 98.6 ± 1.6 | 96.5 ± 1.9 |
| 350 | 97.6 ± 1.1 | 96.1 ± 1.6 | 94.1 ± 2.3 | 91.6 ± 3.6 |
| 400 | 94.8 ± 1.2 | 93.5 ± 1.7 | 87.5 ± 3.7 | 84.7 ± 4.8 |
| 450 | 94.4 ± 1.2 | 92.5 ± 1.6 | 84.9 ± 3.5 | 82.7 ± 3.8 |
| 500 | 92.7 ± 1.1 | 91.0 ± 1.7 | 80.8 ± 2.5 | 78.4 ± 3.3 |
| 550 | 92.3 ± 1.1 | 91.1 ± 2.1 | 79.1 ± 2.6 | 76.5 ± 3.6 |
| 600 | 92.5 ± 1.1 | 91.3 ± 1.4 | 79.7 ± 2.6 | 76.7 ± 3.8 |
| 650 | 90.5 ± 1.1 | 88.9 ± 1.7 | 76.0 ± 2.7 | 73.9 ± 3.1 |
| 700 | 90.4 ± 1.3 | 89.2 ± 1.4 | 76.2 ± 3.9 | 74.1 ± 3.9 |
| 750 | 90.3 ± 0.6 | 88.7 ± 2.0 | 74.3 ± 2.9 | 71.6 ± 3.8 |
| 800 | 90.6 ± 1.1 | 89.8 ± 2.0 | 76.6 ± 2.5 | 74.1 ± 3.2 |
| 850 | 90.3 ± 1.0 | 89.4 ± 1.9 | 75.7 ± 3.9 | 73.3 ± 4.7 |
| 900 | 90.3 ± 1.1 | 89.2 ± 2.1 | 75.1 ± 3.1 | 72.2 ± 3.6 |
| 950 | 89.8 ± 0.9 | 88.5 ± 2.0 | 73.1 ± 2.3 | 71.0 ± 2.6 |
| 1000 | 89.0 ± 1.1 | 87.9 ± 2.1 | 72.9 ± 3.3 | 71.1 ± 3.2 |
| 1050 | 89.2 ± 1.4 | 88.3 ± 1.7 | 72.6 ± 2.5 | 71.3 ± 2.8 |
| 1100 | 89.7 ± 1.3 | 88.4 ± 2.5 | 75.2 ± 3.6 | 72.5 ± 4.8 |
| 1150 | 89.4 ± 1.2 | 87.8 ± 2.0 | 73.1 ± 3.4 | 71.6 ± 3.3 |
| 1200 | 89.2 ± 1.2 | 88.5 ± 2.1 | 73.8 ± 2.5 | 71.7 ± 3.6 |