r7_c7
done · peak unseen_mismatch 99.4
α schedule
α 5.0 ×25 α 5.0 ×25 α 5.0 ×25 α 1.866 ×25 α 1.8518 ×25 α 5.0 ×25 α 5.0 ×25 α 5.0 ×25 α 1.6768 ×25 α 1.5896 ×25 α 1.479 ×25 α 5.0 ×25 α 4.743 ×25 α 0.9761 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.9761
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 ± 1.3 | 50.5 ± 1.6 | 49.9 ± 1.4 | 49.7 ± 2.0 |
| 50 | 51.5 ± 2.0 | 50.5 ± 1.5 | 50.0 ± 1.3 | 50.0 ± 1.6 |
| 75 | 53.4 ± 2.6 | 50.7 ± 1.7 | 50.4 ± 1.3 | 49.6 ± 1.5 |
| 100 | 58.5 ± 9.6 | 55.8 ± 5.6 | 53.5 ± 7.7 | 50.7 ± 3.7 |
| 125 | 71.5 ± 14.6 | 65.1 ± 12.7 | 63.8 ± 14.7 | 58.5 ± 11.0 |
| 150 | 81.9 ± 10.3 | 78.0 ± 10.1 | 72.7 ± 12.3 | 68.2 ± 11.0 |
| 175 | 86.4 ± 4.9 | 83.3 ± 4.9 | 74.5 ± 9.4 | 71.3 ± 8.8 |
| 200 | 84.8 ± 3.8 | 82.7 ± 3.8 | 71.4 ± 7.6 | 68.4 ± 6.6 |
| 225 | 94.5 ± 2.8 | 92.8 ± 2.8 | 86.4 ± 6.3 | 84.0 ± 6.1 |
| 250 | 97.1 ± 1.4 | 95.4 ± 1.6 | 90.2 ± 4.4 | 87.1 ± 4.8 |
| 275 | 98.4 ± 0.7 | 97.4 ± 0.9 | 93.9 ± 3.0 | 91.0 ± 3.3 |
| 300 | 90.7 ± 2.4 | 88.4 ± 2.2 | 78.5 ± 4.2 | 75.2 ± 4.0 |
| 325 | 91.0 ± 2.0 | 88.8 ± 2.1 | 79.5 ± 3.8 | 75.6 ± 4.7 |
| 350 | 100.0 ± 0.1 | 99.8 ± 0.3 | 99.8 ± 0.3 | 99.4 ± 0.7 |