fixed_14_single
done · peak unseen_mismatch 88.3
Referenced by
α schedule
α 1.4 ×1200
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 | 84.8 ± 22.6 | 45.2 ± 19.5 | 84.4 ± 22.9 | 45.3 ± 19.0 |
| 450 | 100.0 ± 0.0 | 36.1 ± 30.8 | 100.0 ± 0.0 | 36.9 ± 30.6 |
| 600 | 100.0 ± 0.0 | 69.7 ± 26.0 | 100.0 ± 0.0 | 69.8 ± 25.7 |
| 750 | 99.9 ± 0.4 | 71.3 ± 29.4 | 99.8 ± 0.7 | 70.8 ± 28.6 |
| 900 | 99.8 ± 0.3 | 80.3 ± 23.3 | 99.5 ± 0.9 | 77.8 ± 21.7 |
| 1050 | 99.8 ± 0.4 | 93.1 ± 7.1 | 99.5 ± 0.5 | 88.3 ± 7.9 |
| 1200 | 99.7 ± 0.3 | 93.3 ± 8.0 | 98.4 ± 1.9 | 88.2 ± 9.1 |