s4_e1_blk
done · peak unseen_mismatch 96.8
α schedule
α 4.0 ×300 α 3.0 ×300 α 2.0 ×300 α 1.0 ×300
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 |
|---|---|---|---|---|
| 50 | 51.7 ± 1.9 | 49.6 ± 1.0 | 49.9 ± 2.1 | 50.5 ± 1.2 |
| 100 | 58.7 ± 7.3 | 56.1 ± 6.1 | 52.6 ± 4.2 | 51.5 ± 2.9 |
| 150 | 76.9 ± 6.3 | 75.5 ± 5.2 | 65.1 ± 6.5 | 61.0 ± 5.8 |
| 200 | 82.0 ± 2.5 | 80.4 ± 1.7 | 64.6 ± 3.8 | 63.1 ± 3.1 |
| 250 | 83.9 ± 2.0 | 82.2 ± 2.0 | 64.0 ± 4.2 | 62.4 ± 3.2 |
| 300 | 85.9 ± 1.6 | 84.5 ± 2.2 | 67.5 ± 4.9 | 66.6 ± 4.4 |
| 350 | 90.9 ± 1.9 | 89.1 ± 1.5 | 75.6 ± 4.2 | 72.6 ± 4.4 |
| 400 | 90.6 ± 1.2 | 88.2 ± 1.8 | 73.2 ± 4.6 | 68.8 ± 3.2 |
| 450 | 91.4 ± 1.5 | 89.9 ± 1.7 | 73.5 ± 5.8 | 69.2 ± 5.6 |
| 500 | 91.8 ± 1.4 | 89.7 ± 1.7 | 75.5 ± 4.7 | 70.3 ± 4.8 |
| 550 | 92.0 ± 1.2 | 90.9 ± 1.8 | 76.5 ± 3.0 | 71.6 ± 4.2 |
| 600 | 92.4 ± 1.3 | 90.7 ± 1.9 | 75.9 ± 3.3 | 71.0 ± 4.7 |
| 650 | 96.4 ± 1.3 | 94.5 ± 1.5 | 84.2 ± 3.7 | 77.1 ± 5.2 |
| 700 | 97.6 ± 0.8 | 96.1 ± 0.8 | 87.5 ± 3.6 | 80.0 ± 4.7 |
| 750 | 97.5 ± 0.8 | 96.2 ± 0.8 | 88.3 ± 3.0 | 81.4 ± 3.4 |
| 800 | 98.3 ± 0.5 | 96.7 ± 1.0 | 89.9 ± 2.9 | 82.1 ± 3.3 |
| 850 | 98.2 ± 0.6 | 96.6 ± 0.9 | 90.2 ± 3.4 | 83.5 ± 4.7 |
| 900 | 98.2 ± 0.4 | 96.5 ± 0.9 | 90.3 ± 2.4 | 82.2 ± 3.6 |
| 950 | 100.0 ± 0.1 | 99.7 ± 0.4 | 99.4 ± 0.6 | 95.6 ± 3.6 |
| 1000 | 100.0 ± 0.0 | 99.8 ± 0.2 | 99.5 ± 0.5 | 96.8 ± 2.2 |
| 1050 | 100.0 ± 0.0 | 99.6 ± 0.3 | 99.4 ± 0.4 | 94.0 ± 3.8 |
| 1100 | 100.0 ± 0.0 | 99.8 ± 0.3 | 99.3 ± 0.7 | 95.0 ± 3.1 |
| 1150 | 100.0 ± 0.1 | 98.8 ± 1.4 | 99.1 ± 0.9 | 88.4 ± 11.0 |
| 1200 | 100.0 ± 0.1 | 99.3 ± 0.8 | 99.0 ± 1.2 | 89.6 ± 10.9 |