r9_c6
done · peak unseen_mismatch 99.6
α schedule
α 4.6813 ×25 α 4.6695 ×25 α 4.6439 ×25 α 1.1525 ×25 α 1.1385 ×25 α 4.4759 ×25 α 4.3879 ×25 α 4.2832 ×25 α 1.0419 ×25 α 1.0071 ×25 α 0.968 ×25 α 3.6927 ×25 α 3.5009 ×25 α 0.8239 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.8239
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 | 50.8 ± 1.3 | 50.6 ± 1.5 | 49.8 ± 1.5 | 49.7 ± 1.9 |
| 50 | 51.5 ± 1.7 | 50.5 ± 1.5 | 50.1 ± 1.3 | 50.0 ± 1.6 |
| 75 | 52.8 ± 2.4 | 50.5 ± 1.8 | 50.5 ± 1.3 | 49.4 ± 1.6 |
| 100 | 55.3 ± 7.0 | 52.7 ± 2.9 | 52.9 ± 5.7 | 50.2 ± 2.2 |
| 125 | 66.1 ± 15.4 | 55.8 ± 7.3 | 62.5 ± 15.6 | 53.0 ± 6.5 |
| 150 | 80.9 ± 16.4 | 74.2 ± 14.3 | 76.9 ± 17.6 | 68.8 ± 13.9 |
| 175 | 88.0 ± 11.8 | 84.2 ± 11.1 | 82.4 ± 13.0 | 77.5 ± 12.7 |
| 200 | 90.1 ± 6.5 | 87.5 ± 7.0 | 82.2 ± 10.0 | 78.6 ± 10.4 |
| 225 | 99.0 ± 1.2 | 96.7 ± 3.7 | 97.9 ± 2.3 | 94.4 ± 5.3 |
| 250 | 99.9 ± 0.1 | 98.4 ± 2.0 | 99.5 ± 0.8 | 96.8 ± 2.9 |
| 275 | 100.0 ± 0.0 | 98.5 ± 3.0 | 99.7 ± 0.8 | 97.4 ± 4.2 |
| 300 | 96.0 ± 2.9 | 93.4 ± 3.6 | 89.5 ± 6.9 | 84.4 ± 8.2 |
| 325 | 94.6 ± 2.2 | 91.8 ± 2.9 | 87.7 ± 5.0 | 82.2 ± 6.1 |
| 350 | 100.0 ± 0.0 | 99.9 ± 0.3 | 100.0 ± 0.0 | 99.6 ± 0.9 |