r9_c5
done · peak unseen_mismatch 95.2
α schedule
α 5.0 ×25 α 5.0 ×25 α 5.0 ×25 α 1.6903 ×25 α 1.6877 ×25 α 5.0 ×25 α 5.0 ×25 α 5.0 ×25 α 1.6545 ×25 α 1.6377 ×25 α 1.6164 ×25 α 5.0 ×25 α 5.0 ×25 α 1.5187 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.5187
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 | 57.7 ± 8.9 | 54.7 ± 4.5 | 53.4 ± 6.8 | 50.5 ± 2.5 |
| 125 | 70.2 ± 14.9 | 62.6 ± 11.7 | 63.8 ± 15.0 | 56.8 ± 9.7 |
| 150 | 81.8 ± 11.3 | 77.4 ± 11.0 | 73.7 ± 13.2 | 68.9 ± 11.9 |
| 175 | 87.1 ± 5.4 | 84.1 ± 5.5 | 76.3 ± 10.0 | 72.1 ± 9.8 |
| 200 | 86.0 ± 4.3 | 83.7 ± 4.3 | 73.2 ± 8.4 | 70.0 ± 8.0 |
| 225 | 95.5 ± 2.6 | 93.9 ± 2.8 | 89.1 ± 5.8 | 86.3 ± 6.2 |
| 250 | 96.7 ± 1.3 | 95.0 ± 1.6 | 89.4 ± 4.7 | 86.1 ± 5.2 |
| 275 | 97.9 ± 0.9 | 96.6 ± 1.3 | 92.0 ± 3.3 | 88.7 ± 4.0 |
| 300 | 91.0 ± 2.1 | 89.0 ± 2.2 | 78.1 ± 5.1 | 74.9 ± 5.4 |
| 325 | 90.8 ± 1.8 | 88.7 ± 1.7 | 78.9 ± 3.8 | 74.9 ± 4.5 |
| 350 | 99.2 ± 0.5 | 98.6 ± 0.8 | 97.0 ± 1.8 | 95.2 ± 2.2 |