fixed_14_fresh
done · peak unseen_mismatch 88.6
Referenced by
α schedule
α 1.4 ×150 α 1.4 ×150 α 1.4 ×150 α 1.4 ×150 α 1.4 ×150 α 1.4 ×150 α 1.4 ×150 α 1.4 ×150
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.4
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 | 52.8 ± 3.6 | 49.1 ± 1.9 | 52.6 ± 3.6 | 47.6 ± 3.0 |
| 300 | 90.1 ± 19.2 | 38.8 ± 20.7 | 89.8 ± 19.6 | 39.0 ± 20.1 |
| 450 | 100.0 ± 0.0 | 46.0 ± 28.5 | 100.0 ± 0.0 | 46.8 ± 28.4 |
| 600 | 100.0 ± 0.0 | 36.1 ± 28.7 | 100.0 ± 0.0 | 37.1 ± 28.5 |
| 750 | 100.0 ± 0.1 | 51.6 ± 32.1 | 100.0 ± 0.0 | 52.4 ± 31.9 |
| 900 | 100.0 ± 0.1 | 74.6 ± 32.6 | 99.8 ± 0.5 | 72.5 ± 31.1 |
| 1050 | 99.9 ± 0.2 | 80.5 ± 26.6 | 99.2 ± 1.3 | 77.3 ± 25.7 |
| 1200 | 99.8 ± 0.3 | 91.5 ± 13.7 | 99.4 ± 0.9 | 88.6 ± 12.4 |