T300
done · peak unseen_mismatch 59.2
α schedule
α 5.0 ×7 α 5.0 ×7 α 5.0 ×7 α 5.0 ×7 α 5.0 ×7 α 5.0 ×7 α 5.0 ×7 α 5.0 ×7 α 0.7 ×7 α 0.7 ×7 α 0.7 ×7 α 0.7 ×7 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 3.1429 ×6 α 3.1429 ×6 α 3.1429 ×6 α 3.1429 ×6 α 3.1429 ×6 α 3.1429 ×6 α 3.1429 ×6 α 3.1429 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 1.2857 ×6 α 1.2857 ×6 α 1.2857 ×6 α 1.2857 ×6 α 1.2857 ×6 α 1.2857 ×6 α 1.2857 ×6 α 1.2857 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6 α 0.7 ×6
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.7
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 | 51.5 ± 1.5 | 49.9 ± 1.3 | 49.8 ± 1.9 | 49.6 ± 2.2 |
| 50 | 50.9 ± 1.8 | 50.0 ± 1.2 | 50.2 ± 1.6 | 50.4 ± 1.5 |
| 75 | 51.9 ± 1.6 | 50.0 ± 1.8 | 50.6 ± 1.8 | 50.2 ± 2.2 |
| 100 | 52.4 ± 2.7 | 49.8 ± 1.7 | 51.6 ± 2.6 | 48.7 ± 1.9 |
| 125 | 60.6 ± 12.5 | 48.3 ± 5.2 | 59.7 ± 12.8 | 48.3 ± 5.0 |
| 150 | 75.9 ± 19.7 | 59.0 ± 14.5 | 73.8 ± 20.3 | 57.9 ± 13.5 |
| 175 | 88.5 ± 14.5 | 60.5 ± 14.5 | 87.0 ± 15.8 | 59.2 ± 14.1 |
| 200 | 96.9 ± 9.8 | 47.5 ± 21.9 | 96.9 ± 9.4 | 46.8 ± 21.1 |
| 225 | 100.0 ± 0.2 | 52.4 ± 16.3 | 100.0 ± 0.1 | 52.3 ± 16.0 |
| 250 | 100.0 ± 0.0 | 54.5 ± 26.8 | 100.0 ± 0.0 | 53.7 ± 26.4 |
| 275 | 100.0 ± 0.0 | 52.6 ± 27.1 | 100.0 ± 0.0 | 52.6 ± 27.1 |
| 300 | 100.0 ± 0.0 | 52.1 ± 25.6 | 100.0 ± 0.0 | 51.8 ± 25.0 |