s5_e0_blk
done · peak unseen_mismatch 94.9
α schedule
α 5.0 ×200 α 4.0 ×200 α 3.0 ×200 α 2.0 ×200 α 1.0 ×200 α 0.0 ×200
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.0
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 | 51.8 ± 2.1 | 49.9 ± 1.2 | 50.2 ± 1.8 | 50.0 ± 1.4 |
| 100 | 59.6 ± 6.4 | 56.9 ± 4.6 | 52.8 ± 3.7 | 51.7 ± 3.0 |
| 150 | 76.5 ± 5.5 | 74.3 ± 5.6 | 62.2 ± 5.9 | 59.1 ± 4.4 |
| 200 | 79.5 ± 1.9 | 78.0 ± 2.0 | 60.5 ± 3.5 | 59.9 ± 3.0 |
| 250 | 82.5 ± 2.4 | 81.1 ± 1.7 | 63.1 ± 3.6 | 61.5 ± 3.6 |
| 300 | 85.3 ± 1.5 | 84.0 ± 1.9 | 67.1 ± 4.4 | 65.2 ± 4.5 |
| 350 | 86.9 ± 2.0 | 85.7 ± 1.9 | 69.1 ± 5.8 | 66.8 ± 5.9 |
| 400 | 86.4 ± 1.6 | 84.8 ± 2.2 | 67.3 ± 1.9 | 64.5 ± 2.8 |
| 450 | 90.3 ± 1.1 | 88.5 ± 2.1 | 72.6 ± 3.1 | 69.8 ± 5.4 |
| 500 | 90.6 ± 1.3 | 89.3 ± 1.5 | 72.6 ± 5.3 | 69.4 ± 5.9 |
| 550 | 91.6 ± 1.6 | 89.2 ± 2.0 | 74.5 ± 5.4 | 70.0 ± 5.0 |
| 600 | 91.9 ± 1.1 | 90.2 ± 1.4 | 74.8 ± 4.5 | 70.7 ± 4.9 |
| 650 | 95.6 ± 1.1 | 94.5 ± 1.4 | 81.9 ± 3.3 | 76.3 ± 4.8 |
| 700 | 97.4 ± 0.7 | 96.0 ± 0.5 | 87.4 ± 3.5 | 81.4 ± 4.8 |
| 750 | 97.4 ± 0.6 | 96.0 ± 1.1 | 87.3 ± 4.5 | 81.3 ± 5.3 |
| 800 | 97.8 ± 0.8 | 96.7 ± 0.8 | 88.1 ± 3.6 | 82.9 ± 4.0 |
| 850 | 100.0 ± 0.1 | 99.5 ± 0.4 | 99.2 ± 0.5 | 94.9 ± 3.8 |
| 900 | 100.0 ± 0.1 | 98.9 ± 2.1 | 98.8 ± 0.9 | 92.4 ± 7.3 |
| 950 | 100.0 ± 0.0 | 99.5 ± 0.6 | 99.4 ± 0.5 | 94.6 ± 4.7 |
| 1000 | 100.0 ± 0.0 | 98.8 ± 1.9 | 99.5 ± 0.5 | 90.5 ± 7.8 |
| 1050 | 100.0 ± 0.0 | 70.1 ± 30.5 | 100.0 ± 0.0 | 66.5 ± 30.8 |
| 1100 | 100.0 ± 0.0 | 59.2 ± 27.0 | 100.0 ± 0.0 | 56.0 ± 25.1 |
| 1150 | 100.0 ± 0.0 | 25.6 ± 27.1 | 99.9 ± 0.1 | 26.7 ± 27.0 |
| 1200 | 100.0 ± 0.0 | 23.6 ± 25.6 | 99.9 ± 0.3 | 25.9 ± 24.3 |