r6_c3
done · peak unseen_mismatch 97.0
α schedule
α 2.9099 ×25 α 2.9091 ×25 α 2.9057 ×25 α 1.1834 ×25 α 1.1779 ×25 α 2.8634 ×25 α 2.8325 ×25 α 2.7901 ×25 α 1.1165 ×25 α 1.0875 ×25 α 1.0515 ×25 α 2.4681 ×25 α 2.3401 ×25 α 0.8939 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.8939
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 ± 1.2 | 50.4 ± 1.7 | 50.1 ± 1.5 | 49.5 ± 1.7 |
| 50 | 51.3 ± 1.5 | 50.3 ± 1.5 | 49.9 ± 1.2 | 49.8 ± 1.7 |
| 75 | 52.0 ± 1.7 | 49.9 ± 1.7 | 50.5 ± 1.4 | 49.4 ± 1.5 |
| 100 | 52.9 ± 5.1 | 50.8 ± 1.6 | 51.4 ± 3.4 | 49.7 ± 1.5 |
| 125 | 59.2 ± 12.4 | 50.3 ± 2.2 | 56.8 ± 12.3 | 49.3 ± 1.8 |
| 150 | 68.5 ± 18.7 | 58.9 ± 10.7 | 66.5 ± 18.9 | 56.4 ± 9.4 |
| 175 | 79.9 ± 18.3 | 72.9 ± 16.6 | 76.7 ± 19.4 | 69.7 ± 15.7 |
| 200 | 87.7 ± 15.1 | 83.1 ± 14.5 | 84.2 ± 16.3 | 78.5 ± 15.1 |
| 225 | 95.3 ± 12.0 | 86.4 ± 17.9 | 94.2 ± 12.9 | 84.9 ± 17.7 |
| 250 | 98.6 ± 5.2 | 88.9 ± 18.0 | 98.1 ± 5.9 | 87.4 ± 17.9 |
| 275 | 100.0 ± 0.0 | 91.2 ± 15.6 | 99.9 ± 0.2 | 90.4 ± 15.5 |
| 300 | 99.0 ± 1.3 | 94.9 ± 8.0 | 97.3 ± 3.1 | 91.2 ± 8.5 |
| 325 | 98.6 ± 1.2 | 95.6 ± 8.6 | 96.4 ± 3.3 | 91.5 ± 8.6 |
| 350 | 100.0 ± 0.0 | 97.7 ± 7.1 | 99.9 ± 0.2 | 97.0 ± 7.7 |