p01
done · peak unseen_mismatch 50.0
α schedule
α 0.1917 ×25 α 0.2406 ×25 α 0.3064 ×25 α 0.3844 ×25 α 0.4723 ×25 α 0.5689 ×25 α 0.673 ×25 α 0.7841 ×25 α 0.9016 ×25 α 1.0249 ×25 α 1.1538 ×25 α 1.2879 ×25 α 1.4269 ×25 α 1.5706 ×25 α 1.7189 ×25 α 1.8715 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.8715
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 | 50.3 ± 1.4 | 49.7 ± 1.5 | 49.8 ± 1.7 | 49.4 ± 1.8 |
| 50 | 50.2 ± 1.7 | 50.0 ± 1.4 | 50.0 ± 1.4 | 50.0 ± 1.7 |
| 75 | 50.5 ± 1.5 | 49.3 ± 1.8 | 50.3 ± 1.4 | 49.0 ± 1.6 |
| 100 | 50.0 ± 2.0 | 49.9 ± 1.6 | 49.7 ± 1.5 | 50.0 ± 1.4 |
| 125 | 51.4 ± 2.1 | 49.4 ± 1.5 | 50.7 ± 1.8 | 49.2 ± 1.8 |
| 150 | 54.2 ± 9.7 | 48.4 ± 4.4 | 53.6 ± 10.0 | 48.1 ± 3.9 |
| 175 | 60.3 ± 17.9 | 44.4 ± 9.2 | 60.2 ± 17.6 | 45.1 ± 9.2 |
| 200 | 64.5 ± 19.9 | 39.0 ± 15.2 | 64.7 ± 19.9 | 39.8 ± 15.7 |
| 225 | 70.9 ± 22.4 | 36.5 ± 17.8 | 71.1 ± 22.2 | 37.3 ± 18.3 |
| 250 | 75.3 ± 23.9 | 35.4 ± 18.8 | 75.5 ± 23.7 | 35.9 ± 18.6 |
| 275 | 80.2 ± 24.2 | 34.1 ± 20.5 | 79.9 ± 24.6 | 34.2 ± 20.4 |
| 300 | 80.8 ± 23.5 | 29.7 ± 20.7 | 81.2 ± 23.0 | 29.8 ± 21.0 |
| 325 | 86.6 ± 19.6 | 32.7 ± 17.4 | 86.4 ± 20.0 | 33.3 ± 17.5 |
| 350 | 90.6 ± 18.5 | 31.9 ± 24.6 | 90.6 ± 18.5 | 32.0 ± 24.8 |
| 375 | 94.7 ± 12.8 | 39.9 ± 25.2 | 94.7 ± 12.7 | 40.2 ± 24.9 |
| 400 | 97.6 ± 10.4 | 38.7 ± 27.9 | 97.6 ± 9.6 | 39.1 ± 27.6 |