s2_e0
done · peak unseen_mismatch 94.0
α schedule
α 2.0 ×50 α 1.913 ×50 α 1.8261 ×50 α 1.7391 ×50 α 1.6522 ×50 α 1.5652 ×50 α 1.4783 ×50 α 1.3913 ×50 α 1.3043 ×50 α 1.2174 ×50 α 1.1304 ×50 α 1.0435 ×50 α 0.9565 ×50 α 0.8696 ×50 α 0.7826 ×50 α 0.6957 ×50 α 0.6087 ×50 α 0.5217 ×50 α 0.4348 ×50 α 0.3478 ×50 α 0.2609 ×50 α 0.1739 ×50 α 0.087 ×50 α 0.0 ×50
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 | 51.2 ± 0.9 | 49.3 ± 0.9 | 50.0 ± 1.5 | 50.5 ± 1.4 |
| 100 | 51.0 ± 2.3 | 49.8 ± 1.5 | 51.0 ± 1.8 | 49.3 ± 1.3 |
| 150 | 60.9 ± 8.2 | 51.0 ± 6.0 | 58.5 ± 7.5 | 51.3 ± 3.9 |
| 200 | 94.7 ± 8.1 | 60.7 ± 21.3 | 93.5 ± 9.8 | 59.0 ± 20.9 |
| 250 | 99.8 ± 0.6 | 54.2 ± 31.2 | 99.6 ± 1.0 | 54.0 ± 30.2 |
| 300 | 99.9 ± 0.2 | 75.7 ± 29.8 | 99.8 ± 0.5 | 75.2 ± 29.4 |
| 350 | 99.9 ± 0.1 | 74.0 ± 30.6 | 99.8 ± 0.3 | 73.2 ± 30.2 |
| 400 | 99.8 ± 0.3 | 90.3 ± 15.4 | 99.7 ± 0.7 | 89.3 ± 14.7 |
| 450 | 99.9 ± 0.1 | 89.6 ± 17.6 | 99.7 ± 0.4 | 88.4 ± 16.8 |
| 500 | 99.9 ± 0.1 | 93.3 ± 12.4 | 99.8 ± 0.3 | 91.2 ± 11.9 |
| 550 | 99.9 ± 0.1 | 95.5 ± 10.0 | 99.5 ± 1.0 | 94.0 ± 9.4 |
| 600 | 100.0 ± 0.0 | 94.7 ± 10.8 | 99.6 ± 0.7 | 92.3 ± 10.2 |
| 650 | 100.0 ± 0.0 | 93.9 ± 14.0 | 99.8 ± 0.3 | 90.9 ± 14.4 |
| 700 | 100.0 ± 0.0 | 96.6 ± 3.6 | 99.4 ± 1.4 | 92.8 ± 6.5 |
| 750 | 100.0 ± 0.0 | 95.5 ± 5.6 | 99.8 ± 0.3 | 91.1 ± 9.0 |
| 800 | 100.0 ± 0.0 | 88.4 ± 15.3 | 100.0 ± 0.1 | 85.0 ± 15.0 |
| 850 | 100.0 ± 0.0 | 89.2 ± 10.6 | 100.0 ± 0.0 | 85.9 ± 12.0 |
| 900 | 100.0 ± 0.0 | 77.0 ± 18.5 | 100.0 ± 0.0 | 72.2 ± 18.9 |
| 950 | 100.0 ± 0.0 | 60.0 ± 30.1 | 99.8 ± 0.6 | 58.5 ± 28.5 |
| 1000 | 100.0 ± 0.0 | 45.0 ± 22.3 | 99.9 ± 0.1 | 48.4 ± 20.3 |
| 1050 | 100.0 ± 0.0 | 26.7 ± 27.0 | 100.0 ± 0.0 | 31.6 ± 27.8 |
| 1100 | 100.0 ± 0.0 | 25.6 ± 18.0 | 100.0 ± 0.1 | 30.6 ± 15.5 |
| 1150 | 100.0 ± 0.0 | 20.6 ± 18.0 | 99.9 ± 0.3 | 25.5 ± 16.9 |
| 1200 | 100.0 ± 0.0 | 23.7 ± 23.8 | 99.7 ± 0.9 | 30.7 ± 24.1 |