r1_c2
done · peak unseen_mismatch 49.9
α schedule
α 2.2261 ×25 α 1.7656 ×25 α 1.4946 ×25 α 1.2923 ×25 α 1.1274 ×25 α 0.9866 ×25 α 0.8628 ×25 α 0.7518 ×25 α 0.6508 ×25 α 0.5578 ×25 α 0.4714 ×25 α 0.3907 ×25 α 0.3147 ×25 α 0.2429 ×25 α 0.1748 ×25 α 0.1098 ×25
Figures
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.1098
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 | 50.7 ± 1.4 | 50.1 ± 1.6 | 49.8 ± 1.4 | 49.6 ± 1.7 |
| 50 | 50.5 ± 1.6 | 50.4 ± 1.6 | 50.0 ± 1.1 | 49.9 ± 1.8 |
| 75 | 51.1 ± 1.7 | 49.5 ± 1.9 | 50.3 ± 1.6 | 49.6 ± 1.6 |
| 100 | 51.2 ± 2.7 | 50.1 ± 1.6 | 50.4 ± 1.7 | 49.7 ± 1.7 |
| 125 | 53.3 ± 4.3 | 48.6 ± 2.4 | 51.8 ± 4.2 | 48.7 ± 2.4 |
| 150 | 59.3 ± 16.0 | 47.6 ± 5.1 | 58.4 ± 16.3 | 47.5 ± 5.5 |
| 175 | 65.3 ± 18.9 | 42.7 ± 12.8 | 64.9 ± 18.8 | 42.5 ± 12.4 |
| 200 | 76.6 ± 23.6 | 42.8 ± 18.9 | 76.2 ± 23.9 | 43.2 ± 19.5 |
| 225 | 81.1 ± 23.4 | 38.9 ± 19.8 | 81.3 ± 23.1 | 39.7 ± 20.4 |
| 250 | 83.6 ± 22.4 | 37.5 ± 20.8 | 83.5 ± 22.2 | 37.8 ± 20.9 |
| 275 | 88.6 ± 18.9 | 35.6 ± 24.4 | 88.1 ± 19.4 | 35.2 ± 24.3 |
| 300 | 93.5 ± 15.5 | 31.4 ± 24.0 | 93.6 ± 15.3 | 31.4 ± 23.7 |
| 325 | 95.3 ± 13.4 | 27.9 ± 22.2 | 95.2 ± 13.6 | 28.2 ± 22.1 |
| 350 | 96.7 ± 11.5 | 24.1 ± 26.0 | 96.5 ± 12.0 | 24.4 ± 26.0 |
| 375 | 99.9 ± 0.7 | 23.6 ± 26.2 | 99.8 ± 1.0 | 23.6 ± 26.2 |
| 400 | 100.0 ± 0.0 | 23.9 ± 23.5 | 100.0 ± 0.0 | 24.4 ± 23.6 |