r8_c7
done · peak unseen_mismatch 91.6
α schedule
α 5.0 ×25 α 5.0 ×25 α 5.0 ×25 α 1.6099 ×25 α 1.6266 ×25 α 5.0 ×25 α 5.0 ×25 α 5.0 ×25 α 1.6742 ×25 α 1.6835 ×25 α 1.692 ×25 α 5.0 ×25 α 5.0 ×25 α 1.7145 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.7145
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 | 57.3 ± 8.6 | 54.5 ± 4.6 | 53.2 ± 6.4 | 50.6 ± 2.6 |
| 125 | 69.4 ± 15.0 | 62.1 ± 11.9 | 63.2 ± 14.9 | 57.0 ± 10.1 |
| 150 | 81.6 ± 11.1 | 77.6 ± 10.8 | 74.0 ± 13.1 | 69.1 ± 12.1 |
| 175 | 86.9 ± 5.3 | 83.9 ± 5.5 | 77.2 ± 10.4 | 72.8 ± 10.2 |
| 200 | 86.6 ± 4.6 | 84.1 ± 4.7 | 74.5 ± 9.3 | 71.8 ± 8.8 |
| 225 | 95.3 ± 2.7 | 93.4 ± 3.1 | 88.6 ± 6.3 | 85.9 ± 7.0 |
| 250 | 96.5 ± 1.5 | 94.7 ± 1.8 | 89.0 ± 4.5 | 84.9 ± 5.1 |
| 275 | 97.6 ± 0.9 | 96.2 ± 1.4 | 91.5 ± 3.7 | 88.1 ± 3.7 |
| 300 | 91.4 ± 2.2 | 89.1 ± 2.1 | 79.3 ± 4.5 | 75.3 ± 4.6 |
| 325 | 90.6 ± 1.6 | 88.4 ± 1.6 | 78.3 ± 3.1 | 74.5 ± 3.4 |
| 350 | 98.5 ± 0.7 | 97.6 ± 0.9 | 94.5 ± 1.8 | 91.6 ± 2.3 |