r2_c2
done · peak unseen_mismatch 50.1
α schedule
α 1.2808 ×25 α 1.2648 ×25 α 1.2502 ×25 α 1.2363 ×25 α 1.223 ×25 α 1.2101 ×25 α 1.1974 ×25 α 1.185 ×25 α 1.1728 ×25 α 1.1608 ×25 α 1.149 ×25 α 1.1373 ×25 α 1.1257 ×25 α 1.1143 ×25 α 1.103 ×25 α 1.0917 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.0917
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 ± 1.4 | 49.9 ± 1.5 | 49.9 ± 1.3 | 49.5 ± 2.0 |
| 50 | 50.7 ± 1.5 | 49.8 ± 1.6 | 50.2 ± 1.5 | 49.7 ± 1.9 |
| 75 | 50.9 ± 1.3 | 49.5 ± 1.6 | 50.3 ± 1.6 | 49.6 ± 1.8 |
| 100 | 50.5 ± 2.3 | 49.9 ± 1.7 | 50.1 ± 1.9 | 50.1 ± 1.5 |
| 125 | 51.3 ± 2.2 | 49.3 ± 2.0 | 50.5 ± 2.2 | 49.3 ± 2.1 |
| 150 | 54.6 ± 11.1 | 47.9 ± 7.2 | 53.5 ± 10.8 | 47.8 ± 6.5 |
| 175 | 59.3 ± 15.8 | 45.3 ± 9.4 | 58.9 ± 15.4 | 45.5 ± 9.7 |
| 200 | 69.4 ± 21.4 | 45.8 ± 15.7 | 69.0 ± 21.7 | 46.0 ± 16.4 |
| 225 | 73.6 ± 23.8 | 44.3 ± 13.1 | 73.8 ± 23.5 | 44.7 ± 14.0 |
| 250 | 76.7 ± 23.0 | 38.5 ± 16.5 | 76.6 ± 22.9 | 38.5 ± 16.6 |
| 275 | 82.3 ± 21.8 | 40.1 ± 15.9 | 81.7 ± 22.6 | 39.8 ± 15.7 |
| 300 | 87.9 ± 17.4 | 33.4 ± 16.9 | 87.7 ± 17.6 | 32.8 ± 16.9 |
| 325 | 97.6 ± 8.8 | 32.0 ± 19.9 | 97.5 ± 8.8 | 31.6 ± 20.6 |
| 350 | 98.9 ± 5.2 | 30.8 ± 16.3 | 98.8 ± 5.9 | 31.0 ± 16.5 |
| 375 | 100.0 ± 0.0 | 39.4 ± 23.4 | 100.0 ± 0.0 | 39.9 ± 23.3 |
| 400 | 100.0 ± 0.0 | 40.4 ± 25.4 | 100.0 ± 0.0 | 40.4 ± 25.5 |