phi3star_ext
done · peak unseen_mismatch 94.2
α schedule
α 3.199 ×25 α 3.155 ×25 α 3.075 ×25 α 2.963 ×25 α 2.82 ×25 α 2.647 ×25 α 2.445 ×25 α 2.216 ×25 α 1.959 ×25 α 1.675 ×25 α 1.365 ×25 α 1.029 ×25 α 0.6671 ×25 α 0.2802 ×25 α 0.2802 ×50
Results
accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 0.2802
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 |
|---|---|---|---|---|
| 100 | 55.8 ± 5.4 | 52.2 ± 4.1 | 52.1 ± 4.9 | 49.2 ± 1.7 |
| 110 | 57.8 ± 7.0 | 53.8 ± 5.1 | 53.5 ± 5.6 | 50.0 ± 2.1 |
| 120 | 61.6 ± 10.3 | 56.6 ± 7.9 | 56.2 ± 9.1 | 52.0 ± 4.7 |
| 130 | 66.6 ± 11.9 | 60.9 ± 9.7 | 60.4 ± 11.4 | 55.5 ± 6.7 |
| 140 | 70.4 ± 13.5 | 65.8 ± 12.2 | 64.0 ± 13.2 | 58.7 ± 10.1 |
| 150 | 74.9 ± 14.0 | 69.5 ± 12.7 | 66.8 ± 14.0 | 62.0 ± 10.9 |
| 160 | 78.7 ± 14.3 | 74.3 ± 13.5 | 69.9 ± 15.2 | 65.1 ± 11.8 |
| 170 | 82.8 ± 11.1 | 77.6 ± 11.1 | 73.6 ± 13.0 | 68.2 ± 11.6 |
| 180 | 87.6 ± 8.8 | 84.9 ± 8.8 | 78.2 ± 12.6 | 73.3 ± 11.1 |
| 190 | 89.4 ± 8.1 | 86.3 ± 8.5 | 80.5 ± 10.1 | 75.5 ± 8.7 |
| 200 | 90.9 ± 6.5 | 87.7 ± 6.9 | 81.4 ± 8.8 | 77.0 ± 7.8 |
| 210 | 91.7 ± 6.7 | 89.2 ± 7.5 | 83.1 ± 8.9 | 78.7 ± 8.2 |
| 220 | 93.5 ± 3.8 | 91.8 ± 4.1 | 84.9 ± 7.1 | 80.9 ± 7.4 |
| 230 | 94.2 ± 3.9 | 92.8 ± 4.3 | 85.7 ± 5.9 | 81.8 ± 6.4 |
| 240 | 95.9 ± 2.4 | 94.4 ± 2.7 | 88.2 ± 4.8 | 85.0 ± 5.6 |
| 250 | 96.5 ± 1.3 | 94.6 ± 2.6 | 88.7 ± 4.1 | 85.5 ± 4.2 |
| 260 | 98.2 ± 0.7 | 96.4 ± 1.4 | 91.8 ± 3.1 | 88.0 ± 3.0 |
| 270 | 98.5 ± 0.9 | 96.5 ± 2.1 | 93.4 ± 3.6 | 89.0 ± 4.0 |
| 280 | 98.8 ± 0.4 | 96.5 ± 2.5 | 95.4 ± 2.8 | 89.3 ± 3.0 |
| 290 | 99.5 ± 0.2 | 98.3 ± 1.3 | 97.7 ± 1.1 | 94.2 ± 1.8 |
| 300 | 99.9 ± 0.0 | 97.4 ± 0.0 | 99.5 ± 0.0 | 93.8 ± 0.0 |