r3_c1
done · peak unseen_mismatch 96.9
α schedule
α 3.7417 ×25 α 3.7395 ×25 α 3.7306 ×25 α 3.7098 ×25 α 3.6717 ×25 α 3.6105 ×25 α 3.5205 ×25 α 3.3957 ×25 α 3.2299 ×25 α 3.0171 ×25 α 2.7509 ×25 α 2.425 ×25 α 2.0329 ×25 α 1.5681 ×25 α 1.024 ×25 α 0.3941 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.3941
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 ± 1.3 | 50.3 ± 1.7 | 49.9 ± 1.6 | 49.5 ± 1.9 |
| 50 | 51.4 ± 1.6 | 50.3 ± 1.5 | 50.0 ± 1.2 | 49.8 ± 1.8 |
| 75 | 52.1 ± 2.0 | 50.0 ± 1.6 | 50.6 ± 1.3 | 49.4 ± 1.6 |
| 100 | 58.0 ± 7.5 | 55.4 ± 5.0 | 52.6 ± 4.0 | 50.5 ± 2.0 |
| 125 | 69.2 ± 10.5 | 66.2 ± 9.4 | 59.4 ± 9.3 | 56.6 ± 7.8 |
| 150 | 78.3 ± 8.0 | 76.0 ± 7.9 | 66.0 ± 8.2 | 62.9 ± 7.3 |
| 175 | 83.7 ± 5.0 | 81.8 ± 4.3 | 68.5 ± 8.7 | 65.7 ± 8.1 |
| 200 | 84.6 ± 3.2 | 82.7 ± 3.2 | 68.0 ± 6.0 | 65.6 ± 6.3 |
| 225 | 86.2 ± 2.9 | 84.0 ± 2.5 | 69.4 ± 5.6 | 66.9 ± 5.7 |
| 250 | 88.0 ± 2.3 | 86.0 ± 2.3 | 71.2 ± 4.6 | 68.6 ± 4.9 |
| 275 | 89.9 ± 2.4 | 88.0 ± 2.0 | 73.4 ± 5.0 | 71.1 ± 4.9 |
| 300 | 92.1 ± 1.8 | 90.6 ± 2.0 | 77.5 ± 5.1 | 74.7 ± 4.7 |
| 325 | 95.2 ± 1.3 | 93.5 ± 1.3 | 82.7 ± 3.9 | 79.3 ± 4.0 |
| 350 | 98.2 ± 0.7 | 97.2 ± 0.8 | 91.4 ± 2.9 | 88.0 ± 3.8 |
| 375 | 99.8 ± 0.2 | 99.5 ± 0.4 | 98.6 ± 0.9 | 96.9 ± 1.5 |
| 400 | 100.0 ± 0.0 | 97.9 ± 6.4 | 100.0 ± 0.1 | 96.5 ± 9.2 |