p03
done · peak unseen_mismatch 90.4
α schedule
α 4.0412 ×25 α 4.0407 ×25 α 4.038 ×25 α 4.0303 ×25 α 4.0138 ×25 α 3.9841 ×25 α 3.9357 ×25 α 3.8628 ×25 α 3.7588 ×25 α 3.6164 ×25 α 3.4279 ×25 α 3.1848 ×25 α 2.8782 ×25 α 2.4986 ×25 α 2.0361 ×25 α 1.48 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.48
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 ± 1.2 | 50.4 ± 1.6 | 49.9 ± 1.5 | 49.5 ± 1.8 |
| 50 | 51.4 ± 1.6 | 50.2 ± 1.7 | 50.0 ± 1.1 | 49.8 ± 1.6 |
| 75 | 52.6 ± 2.1 | 50.1 ± 1.7 | 50.9 ± 1.3 | 49.4 ± 1.6 |
| 100 | 58.9 ± 7.9 | 56.6 ± 4.9 | 53.4 ± 4.3 | 51.3 ± 2.9 |
| 125 | 70.9 ± 10.4 | 67.8 ± 9.9 | 58.8 ± 9.1 | 56.6 ± 8.1 |
| 150 | 78.4 ± 6.3 | 76.1 ± 6.5 | 64.4 ± 7.8 | 61.7 ± 7.6 |
| 175 | 81.9 ± 3.9 | 79.9 ± 3.5 | 66.1 ± 7.3 | 63.7 ± 7.1 |
| 200 | 83.0 ± 3.1 | 81.1 ± 2.9 | 65.2 ± 5.6 | 63.1 ± 5.7 |
| 225 | 84.1 ± 2.9 | 82.1 ± 2.7 | 66.4 ± 5.1 | 63.9 ± 5.3 |
| 250 | 84.6 ± 2.0 | 83.3 ± 2.0 | 66.2 ± 3.8 | 64.5 ± 4.1 |
| 275 | 86.1 ± 2.3 | 84.6 ± 2.1 | 68.1 ± 4.6 | 66.0 ± 4.4 |
| 300 | 88.6 ± 2.7 | 87.1 ± 2.4 | 72.3 ± 5.8 | 69.7 ± 5.6 |
| 325 | 90.3 ± 2.0 | 88.6 ± 1.4 | 73.4 ± 5.4 | 70.9 ± 5.2 |
| 350 | 92.7 ± 1.6 | 91.2 ± 1.5 | 77.3 ± 4.4 | 74.2 ± 4.6 |
| 375 | 95.2 ± 1.4 | 94.0 ± 1.3 | 82.1 ± 4.4 | 79.4 ± 4.1 |
| 400 | 98.8 ± 0.5 | 97.9 ± 0.8 | 93.6 ± 2.6 | 90.4 ± 2.9 |