s0_e4_blk
done · peak unseen_mismatch 90.6
α schedule
α 0.0 ×240 α 1.0 ×240 α 2.0 ×240 α 3.0 ×240 α 4.0 ×240
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 4.0
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 | 50.4 ± 1.6 | 49.9 ± 1.1 | 49.8 ± 1.9 | 49.6 ± 1.9 |
| 100 | 50.3 ± 1.4 | 50.1 ± 2.2 | 49.8 ± 1.5 | 50.0 ± 1.9 |
| 150 | 52.9 ± 4.9 | 48.8 ± 3.2 | 53.3 ± 5.4 | 47.9 ± 2.6 |
| 200 | 57.3 ± 16.0 | 44.2 ± 13.4 | 57.9 ± 15.6 | 45.0 ± 14.2 |
| 250 | 68.3 ± 22.6 | 45.9 ± 11.8 | 68.7 ± 22.0 | 45.7 ± 11.8 |
| 300 | 82.6 ± 21.3 | 34.0 ± 22.2 | 81.5 ± 22.8 | 34.6 ± 22.3 |
| 350 | 95.0 ± 14.8 | 17.7 ± 16.1 | 95.2 ± 14.2 | 18.1 ± 17.1 |
| 400 | 95.5 ± 13.5 | 16.5 ± 19.0 | 95.3 ± 14.0 | 16.9 ± 19.3 |
| 450 | 100.0 ± 0.0 | 21.4 ± 19.8 | 100.0 ± 0.0 | 21.8 ± 20.4 |
| 500 | 100.0 ± 0.0 | 16.8 ± 12.3 | 100.0 ± 0.0 | 17.3 ± 12.2 |
| 550 | 100.0 ± 0.0 | 36.0 ± 39.6 | 100.0 ± 0.0 | 35.6 ± 39.8 |
| 600 | 100.0 ± 0.0 | 52.3 ± 37.1 | 100.0 ± 0.0 | 51.9 ± 36.6 |
| 650 | 100.0 ± 0.0 | 74.2 ± 34.3 | 100.0 ± 0.0 | 74.1 ± 34.6 |
| 700 | 100.0 ± 0.0 | 82.6 ± 26.2 | 99.9 ± 0.2 | 83.4 ± 24.4 |
| 750 | 99.7 ± 0.3 | 90.4 ± 22.8 | 99.5 ± 0.7 | 89.3 ± 22.8 |
| 800 | 98.9 ± 1.1 | 93.0 ± 13.5 | 98.4 ± 1.3 | 90.6 ± 13.5 |
| 850 | 98.1 ± 1.7 | 91.7 ± 16.8 | 96.1 ± 2.9 | 87.9 ± 18.4 |
| 900 | 97.2 ± 1.8 | 94.1 ± 6.7 | 93.9 ± 3.1 | 88.2 ± 9.3 |
| 950 | 96.5 ± 1.4 | 94.5 ± 2.4 | 92.0 ± 2.8 | 88.9 ± 3.3 |
| 1000 | 94.8 ± 1.8 | 91.2 ± 5.9 | 88.0 ± 2.7 | 83.4 ± 4.3 |
| 1050 | 93.1 ± 1.8 | 91.6 ± 1.4 | 84.5 ± 3.2 | 81.5 ± 3.4 |
| 1100 | 91.2 ± 1.8 | 89.5 ± 1.6 | 80.1 ± 3.1 | 76.4 ± 2.0 |
| 1150 | 90.2 ± 1.6 | 88.6 ± 1.7 | 76.8 ± 3.6 | 74.5 ± 2.4 |
| 1200 | 89.7 ± 1.6 | 88.2 ± 1.8 | 75.6 ± 3.9 | 72.4 ± 4.0 |