r9_c7
done · peak unseen_mismatch 98.5
α schedule
α 3.9283 ×25 α 3.934 ×25 α 3.9553 ×25 α 0.4894 ×25 α 0.4998 ×25 α 4.2196 ×25 α 4.41 ×25 α 4.6692 ×25 α 0.6123 ×25 α 0.6648 ×25 α 0.7296 ×25 α 5.0 ×25 α 5.0 ×25 α 1.0113 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.0113
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 |
|---|---|---|---|---|
| 25 | 50.9 ± 1.2 | 50.3 ± 1.7 | 49.7 ± 1.5 | 49.6 ± 1.7 |
| 50 | 51.4 ± 1.6 | 50.2 ± 1.7 | 50.0 ± 1.1 | 49.8 ± 1.7 |
| 75 | 52.4 ± 2.2 | 50.1 ± 1.6 | 50.8 ± 1.3 | 49.4 ± 1.6 |
| 100 | 53.3 ± 4.7 | 50.8 ± 1.9 | 51.5 ± 3.0 | 49.7 ± 1.6 |
| 125 | 59.9 ± 13.7 | 48.4 ± 5.2 | 58.0 ± 14.1 | 48.2 ± 5.4 |
| 150 | 71.4 ± 18.5 | 57.0 ± 8.9 | 69.7 ± 19.1 | 54.8 ± 7.8 |
| 175 | 84.7 ± 14.6 | 74.2 ± 13.4 | 81.7 ± 15.9 | 70.3 ± 13.6 |
| 200 | 89.8 ± 10.8 | 86.4 ± 11.1 | 86.3 ± 11.9 | 81.5 ± 11.7 |
| 225 | 98.1 ± 5.9 | 90.4 ± 13.8 | 97.7 ± 6.4 | 89.9 ± 13.6 |
| 250 | 100.0 ± 0.1 | 88.8 ± 15.3 | 99.9 ± 0.3 | 88.0 ± 15.7 |
| 275 | 100.0 ± 0.0 | 84.0 ± 22.7 | 100.0 ± 0.0 | 84.0 ± 22.6 |
| 300 | 99.3 ± 1.7 | 96.1 ± 6.2 | 98.2 ± 4.4 | 93.6 ± 8.5 |
| 325 | 97.4 ± 2.6 | 94.3 ± 4.6 | 94.3 ± 5.1 | 89.1 ± 8.3 |
| 350 | 100.0 ± 0.0 | 99.2 ± 1.9 | 100.0 ± 0.1 | 98.5 ± 2.8 |