r9_c4
done · peak unseen_mismatch 98.2
α schedule
α 5.0 ×25 α 5.0 ×25 α 5.0 ×25 α 0.0144 ×25 α 0.0142 ×25 α 5.0 ×25 α 5.0 ×25 α 5.0 ×25 α 0.013 ×25 α 0.0126 ×25 α 0.0121 ×25 α 5.0 ×25 α 5.0 ×25 α 0.0105 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.0105
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 ± 1.3 | 50.5 ± 1.6 | 49.9 ± 1.4 | 49.7 ± 2.0 |
| 50 | 51.5 ± 2.0 | 50.5 ± 1.5 | 50.0 ± 1.3 | 50.0 ± 1.6 |
| 75 | 53.4 ± 2.6 | 50.7 ± 1.7 | 50.4 ± 1.3 | 49.6 ± 1.5 |
| 100 | 54.3 ± 7.5 | 51.1 ± 2.1 | 52.8 ± 6.6 | 50.0 ± 2.2 |
| 125 | 60.7 ± 15.1 | 48.7 ± 5.0 | 58.5 ± 15.6 | 47.7 ± 5.6 |
| 150 | 74.5 ± 19.7 | 57.5 ± 13.0 | 72.9 ± 20.2 | 55.3 ± 11.5 |
| 175 | 85.9 ± 13.1 | 76.0 ± 14.2 | 82.4 ± 14.9 | 72.2 ± 13.6 |
| 200 | 91.2 ± 9.2 | 87.5 ± 9.5 | 86.9 ± 11.5 | 82.2 ± 11.3 |
| 225 | 99.4 ± 1.7 | 90.9 ± 14.4 | 99.3 ± 1.8 | 90.0 ± 15.0 |
| 250 | 100.0 ± 0.0 | 81.8 ± 22.2 | 100.0 ± 0.2 | 80.7 ± 22.9 |
| 275 | 100.0 ± 0.0 | 77.9 ± 21.1 | 100.0 ± 0.0 | 77.6 ± 20.8 |
| 300 | 99.8 ± 1.0 | 95.3 ± 8.2 | 99.7 ± 1.4 | 94.7 ± 8.9 |
| 325 | 98.6 ± 2.9 | 95.1 ± 9.6 | 97.2 ± 5.0 | 92.8 ± 10.6 |
| 350 | 100.0 ± 0.0 | 98.2 ± 6.5 | 100.0 ± 0.0 | 98.2 ± 6.3 |