p09
done · peak unseen_mismatch 50.0
α schedule
α 0.7152 ×25 α 0.7249 ×25 α 0.7324 ×25 α 0.739 ×25 α 0.745 ×25 α 0.7506 ×25 α 0.7559 ×25 α 0.7609 ×25 α 0.7656 ×25 α 0.7702 ×25 α 0.7746 ×25 α 0.7789 ×25 α 0.7831 ×25 α 0.7872 ×25 α 0.7911 ×25 α 0.795 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.795
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 | 50.4 ± 1.3 | 50.1 ± 1.5 | 50.0 ± 1.4 | 49.2 ± 1.7 |
| 50 | 50.5 ± 1.4 | 50.1 ± 1.7 | 50.1 ± 1.4 | 49.9 ± 1.8 |
| 75 | 50.4 ± 1.6 | 49.3 ± 1.8 | 50.0 ± 1.4 | 49.5 ± 1.9 |
| 100 | 50.2 ± 1.7 | 50.1 ± 1.5 | 49.8 ± 1.1 | 50.0 ± 1.5 |
| 125 | 51.5 ± 1.6 | 49.2 ± 1.5 | 50.5 ± 2.1 | 49.0 ± 2.2 |
| 150 | 52.4 ± 5.0 | 49.1 ± 2.4 | 51.5 ± 4.6 | 49.1 ± 1.9 |
| 175 | 57.3 ± 13.8 | 45.3 ± 7.8 | 57.6 ± 13.6 | 46.0 ± 8.0 |
| 200 | 65.3 ± 21.3 | 39.8 ± 16.0 | 65.1 ± 21.2 | 40.7 ± 15.9 |
| 225 | 70.0 ± 22.1 | 36.8 ± 17.6 | 70.3 ± 21.9 | 38.1 ± 18.2 |
| 250 | 76.2 ± 24.3 | 34.1 ± 18.6 | 76.1 ± 24.1 | 34.7 ± 19.1 |
| 275 | 78.0 ± 23.5 | 32.7 ± 22.1 | 77.5 ± 24.0 | 32.0 ± 21.9 |
| 300 | 82.5 ± 22.7 | 29.0 ± 17.9 | 82.8 ± 22.5 | 29.8 ± 18.2 |
| 325 | 87.6 ± 20.1 | 29.2 ± 19.7 | 87.3 ± 20.6 | 29.5 ± 19.4 |
| 350 | 90.7 ± 17.7 | 26.0 ± 20.5 | 90.7 ± 17.6 | 26.2 ± 20.4 |
| 375 | 93.9 ± 15.5 | 25.4 ± 20.9 | 93.8 ± 15.6 | 25.7 ± 20.6 |
| 400 | 96.9 ± 9.6 | 24.0 ± 20.3 | 97.1 ± 9.0 | 24.6 ± 20.6 |