reverse_0_3
done · peak unseen_mismatch 86.9
Referenced by
α schedule
α 0.0 ×150 α 0.4286 ×150 α 0.8571 ×150 α 1.2857 ×150 α 1.7143 ×150 α 2.1429 ×150 α 2.5714 ×150 α 3.0 ×150
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 3.0
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 |
|---|---|---|---|---|
| 150 | 53.2 ± 4.8 | 49.9 ± 3.3 | 52.7 ± 6.2 | 49.1 ± 3.5 |
| 300 | 89.8 ± 20.4 | 21.6 ± 17.4 | 90.1 ± 19.7 | 22.9 ± 18.3 |
| 450 | 95.2 ± 14.3 | 28.8 ± 22.7 | 95.3 ± 14.0 | 29.2 ± 23.1 |
| 600 | 100.0 ± 0.0 | 50.1 ± 33.2 | 100.0 ± 0.0 | 49.4 ± 32.6 |
| 750 | 100.0 ± 0.0 | 53.6 ± 33.9 | 100.0 ± 0.0 | 54.0 ± 33.6 |
| 900 | 100.0 ± 0.0 | 73.5 ± 28.4 | 99.8 ± 0.5 | 71.8 ± 27.8 |
| 1050 | 98.9 ± 1.1 | 89.5 ± 12.8 | 97.8 ± 2.3 | 86.9 ± 13.5 |
| 1200 | 96.3 ± 1.6 | 92.6 ± 4.3 | 92.4 ± 4.6 | 85.4 ± 6.2 |