r4_c1
done · peak unseen_mismatch 93.9
α schedule
α 1.7748 ×25 α 1.7851 ×25 α 1.8009 ×25 α 1.821 ×25 α 1.845 ×25 α 1.8725 ×25 α 1.9031 ×25 α 1.9368 ×25 α 1.9733 ×25 α 2.0125 ×25 α 2.0543 ×25 α 2.0985 ×25 α 2.1451 ×25 α 2.1941 ×25 α 2.2453 ×25 α 2.2986 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 2.2986
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 | 50.7 ± 1.5 | 50.0 ± 1.5 | 50.2 ± 1.5 | 49.4 ± 1.8 |
| 50 | 50.8 ± 1.9 | 50.3 ± 1.5 | 50.0 ± 1.4 | 50.0 ± 1.4 |
| 75 | 50.6 ± 1.5 | 49.7 ± 1.8 | 50.3 ± 1.4 | 49.3 ± 1.7 |
| 100 | 50.8 ± 2.4 | 50.4 ± 1.4 | 50.3 ± 1.6 | 49.8 ± 1.6 |
| 125 | 53.1 ± 3.3 | 49.5 ± 1.7 | 51.5 ± 3.0 | 49.0 ± 2.0 |
| 150 | 58.7 ± 13.1 | 51.3 ± 5.5 | 56.9 ± 12.4 | 50.2 ± 5.5 |
| 175 | 68.5 ± 18.4 | 55.4 ± 11.5 | 67.2 ± 18.3 | 54.8 ± 10.5 |
| 200 | 79.5 ± 21.9 | 62.4 ± 12.4 | 78.4 ± 22.1 | 61.7 ± 12.5 |
| 225 | 83.8 ± 20.5 | 69.9 ± 18.7 | 82.5 ± 21.4 | 68.5 ± 18.7 |
| 250 | 91.0 ± 13.4 | 81.0 ± 16.1 | 88.6 ± 15.2 | 78.6 ± 16.2 |
| 275 | 97.6 ± 5.1 | 90.4 ± 9.8 | 96.6 ± 5.4 | 88.2 ± 11.1 |
| 300 | 99.0 ± 1.6 | 94.6 ± 6.1 | 97.7 ± 3.0 | 91.8 ± 6.9 |
| 325 | 99.1 ± 1.0 | 96.5 ± 4.0 | 97.5 ± 3.1 | 93.9 ± 5.4 |
| 350 | 98.7 ± 1.2 | 96.1 ± 5.0 | 95.9 ± 4.3 | 92.3 ± 6.8 |
| 375 | 98.3 ± 1.4 | 97.0 ± 2.4 | 95.0 ± 4.1 | 92.1 ± 6.1 |
| 400 | 98.0 ± 1.5 | 97.0 ± 1.9 | 94.6 ± 4.0 | 92.2 ± 5.0 |