r8_c2
done · peak unseen_mismatch 96.5
α schedule
α 5.0 ×25 α 5.0 ×25 α 5.0 ×25 α 0.0078 ×25 α 0.0079 ×25 α 5.0 ×25 α 5.0 ×25 α 5.0 ×25 α 0.0083 ×25 α 0.0083 ×25 α 0.0084 ×25 α 5.0 ×25 α 5.0 ×25 α 0.0086 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.0086
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 | 54.5 ± 7.5 | 51.1 ± 2.0 | 52.8 ± 6.3 | 50.0 ± 2.2 |
| 125 | 60.8 ± 15.3 | 48.1 ± 6.7 | 58.7 ± 15.8 | 47.4 ± 7.1 |
| 150 | 75.1 ± 19.1 | 57.8 ± 13.7 | 73.4 ± 19.6 | 55.3 ± 12.4 |
| 175 | 86.2 ± 13.8 | 77.6 ± 14.6 | 83.7 ± 14.7 | 74.0 ± 14.4 |
| 200 | 91.4 ± 8.9 | 87.7 ± 9.4 | 86.9 ± 11.1 | 82.7 ± 11.3 |
| 225 | 99.6 ± 1.2 | 90.3 ± 14.5 | 99.3 ± 1.8 | 89.5 ± 14.9 |
| 250 | 100.0 ± 0.2 | 80.7 ± 27.0 | 99.9 ± 0.3 | 79.9 ± 27.4 |
| 275 | 100.0 ± 0.0 | 69.3 ± 27.7 | 100.0 ± 0.0 | 69.5 ± 27.7 |
| 300 | 100.0 ± 0.1 | 93.7 ± 11.4 | 100.0 ± 0.2 | 93.1 ± 11.5 |
| 325 | 98.7 ± 2.3 | 96.9 ± 3.4 | 97.4 ± 4.0 | 94.4 ± 5.8 |
| 350 | 100.0 ± 0.0 | 96.5 ± 8.1 | 100.0 ± 0.0 | 96.5 ± 8.2 |