r9_c3
done · peak unseen_mismatch 99.9
α schedule
α 4.7584 ×25 α 4.7582 ×25 α 4.7569 ×25 α 0.6878 ×25 α 0.6864 ×25 α 4.7243 ×25 α 4.692 ×25 α 4.641 ×25 α 0.6606 ×25 α 0.645 ×25 α 0.6236 ×25 α 4.1133 ×25 α 3.859 ×25 α 0.5117 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.5117
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 ± 1.2 | 50.5 ± 1.5 | 49.8 ± 1.6 | 49.7 ± 1.9 |
| 50 | 51.5 ± 1.9 | 50.4 ± 1.5 | 50.0 ± 1.3 | 50.0 ± 1.5 |
| 75 | 53.0 ± 2.4 | 50.5 ± 1.6 | 50.4 ± 1.3 | 49.5 ± 1.5 |
| 100 | 54.1 ± 5.7 | 51.2 ± 1.6 | 52.4 ± 4.5 | 49.8 ± 1.8 |
| 125 | 62.0 ± 15.1 | 51.1 ± 4.8 | 59.4 ± 15.7 | 49.7 ± 4.8 |
| 150 | 76.4 ± 18.9 | 67.6 ± 15.1 | 74.4 ± 19.3 | 64.4 ± 14.0 |
| 175 | 86.7 ± 11.8 | 81.0 ± 12.1 | 82.8 ± 13.7 | 76.4 ± 13.0 |
| 200 | 90.6 ± 7.1 | 87.5 ± 7.4 | 84.8 ± 9.9 | 79.9 ± 10.2 |
| 225 | 99.5 ± 1.0 | 91.7 ± 16.4 | 99.1 ± 1.5 | 90.9 ± 16.2 |
| 250 | 100.0 ± 0.2 | 92.4 ± 14.5 | 99.9 ± 0.4 | 91.6 ± 15.3 |
| 275 | 100.0 ± 0.0 | 92.4 ± 12.1 | 100.0 ± 0.0 | 91.6 ± 13.1 |
| 300 | 98.3 ± 2.9 | 96.7 ± 4.3 | 95.7 ± 7.2 | 92.7 ± 9.3 |
| 325 | 96.9 ± 2.7 | 94.4 ± 3.6 | 93.0 ± 5.5 | 88.7 ± 7.7 |
| 350 | 100.0 ± 0.0 | 100.0 ± 0.1 | 100.0 ± 0.0 | 99.9 ± 0.4 |