r7_c6
done · peak unseen_mismatch 95.7
α schedule
α 3.4059 ×25 α 3.4046 ×25 α 3.3996 ×25 α 0.7756 ×25 α 0.7709 ×25 α 3.3343 ×25 α 3.2861 ×25 α 3.2196 ×25 α 0.7169 ×25 α 0.6913 ×25 α 0.6593 ×25 α 2.7095 ×25 α 2.5053 ×25 α 0.5183 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.5183
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 ± 1.5 | 50.3 ± 1.6 | 50.0 ± 1.4 | 49.5 ± 1.8 |
| 50 | 51.5 ± 1.4 | 50.5 ± 1.5 | 50.0 ± 1.4 | 49.9 ± 1.9 |
| 75 | 52.3 ± 2.0 | 50.0 ± 1.7 | 50.4 ± 1.7 | 49.5 ± 1.5 |
| 100 | 53.0 ± 4.4 | 50.5 ± 1.8 | 51.8 ± 4.1 | 49.5 ± 1.7 |
| 125 | 59.5 ± 13.1 | 48.7 ± 4.6 | 56.7 ± 13.3 | 47.6 ± 5.0 |
| 150 | 69.9 ± 18.4 | 56.9 ± 9.9 | 67.8 ± 18.4 | 54.7 ± 8.2 |
| 175 | 81.3 ± 16.8 | 71.7 ± 15.4 | 78.3 ± 17.7 | 68.3 ± 14.5 |
| 200 | 90.7 ± 9.6 | 85.5 ± 10.5 | 86.7 ± 12.3 | 80.7 ± 12.2 |
| 225 | 98.8 ± 3.3 | 86.6 ± 14.3 | 98.2 ± 4.2 | 84.9 ± 14.5 |
| 250 | 100.0 ± 0.2 | 82.8 ± 18.4 | 99.9 ± 0.6 | 82.3 ± 18.3 |
| 275 | 100.0 ± 0.0 | 78.4 ± 25.9 | 100.0 ± 0.0 | 78.2 ± 25.9 |
| 300 | 99.6 ± 1.3 | 86.4 ± 21.7 | 99.1 ± 2.7 | 85.4 ± 21.1 |
| 325 | 99.1 ± 1.3 | 93.8 ± 11.8 | 97.7 ± 3.3 | 90.8 ± 11.5 |
| 350 | 100.0 ± 0.0 | 96.0 ± 11.0 | 100.0 ± 0.0 | 95.7 ± 10.7 |