imagining syntax

← refine_v1

smooth_K48_cont

done · peak unseen_mismatch 91.3

Referenced by

α schedule

α 2.9821 ×19 α 2.9464 ×19 α 2.9107 ×19 α 2.875 ×19 α 2.8393 ×19 α 2.8036 ×19 α 2.7679 ×19 α 2.7321 ×19 α 2.6964 ×19 α 2.6607 ×19 α 2.625 ×19 α 2.5893 ×19 α 2.5536 ×19 α 2.5179 ×19 α 2.4821 ×19 α 2.4464 ×19 α 2.4107 ×19 α 2.375 ×19 α 2.3393 ×19 α 2.3036 ×19 α 2.2679 ×19 α 2.2321 ×19 α 2.1964 ×19 α 2.1607 ×19 α 2.125 ×19 α 2.0893 ×19 α 2.0536 ×19 α 2.0178 ×19 α 1.9821 ×19 α 1.9464 ×19 α 1.9107 ×19 α 1.875 ×19 α 1.8393 ×19 α 1.8036 ×19 α 1.7678 ×19 α 1.7321 ×19 α 1.6964 ×18 α 1.6607 ×18 α 1.625 ×18 α 1.5893 ×18 α 1.5536 ×18 α 1.5178 ×18 α 1.4821 ×18 α 1.4464 ×18 α 1.4107 ×18 α 1.375 ×18 α 1.3393 ×18 α 1.3036 ×18

direction
descending
α range
1.3036–2.9821
stages
48

Figures

schedule.png
schedule.png
trajectory.png
trajectory.png

Results

accuracy by training iteration, mean ± sd over seeds · schedule ends at α = 1.3036

seen_* probed at reference α = 1.4 · unseen_* = held-out pairings, uniform (α-independent) · 1000 pairs per condition

iterationseen_matchseen_mismatchunseen_matchunseen_mismatch
25 51.4 ± 1.549.5 ± 1.649.9 ± 1.849.7 ± 2.0
50 50.9 ± 1.749.5 ± 1.450.0 ± 1.550.8 ± 1.3
75 52.0 ± 1.950.0 ± 1.850.5 ± 1.750.1 ± 1.9
100 52.9 ± 3.051.7 ± 2.651.2 ± 1.649.8 ± 1.2
125 64.2 ± 11.158.5 ± 7.758.6 ± 10.054.2 ± 4.9
150 79.0 ± 12.871.5 ± 13.072.0 ± 14.464.8 ± 12.5
175 87.7 ± 10.681.9 ± 11.080.5 ± 12.573.8 ± 11.5
200 90.5 ± 9.786.3 ± 9.082.8 ± 10.677.1 ± 9.6
225 91.5 ± 8.689.8 ± 8.183.2 ± 9.979.6 ± 9.9
250 92.5 ± 4.490.4 ± 5.282.8 ± 6.478.9 ± 6.8
275 93.9 ± 2.492.0 ± 2.683.5 ± 6.379.6 ± 6.3
300 94.2 ± 1.592.3 ± 1.884.0 ± 5.279.8 ± 5.4
325 94.8 ± 1.692.7 ± 2.084.0 ± 4.778.9 ± 5.4
350 94.7 ± 1.292.8 ± 1.583.1 ± 3.678.0 ± 4.2
375 94.9 ± 1.293.1 ± 1.884.4 ± 3.479.8 ± 4.7
400 95.8 ± 1.393.8 ± 1.584.8 ± 2.679.8 ± 3.9
450 96.0 ± 1.294.3 ± 1.486.0 ± 3.980.5 ± 4.9
500 96.8 ± 0.794.8 ± 1.288.2 ± 2.581.0 ± 4.6
550 97.8 ± 0.795.8 ± 0.989.9 ± 2.384.1 ± 3.3
600 98.1 ± 0.696.6 ± 1.092.0 ± 2.185.2 ± 3.7
650 98.5 ± 0.497.1 ± 0.892.3 ± 2.784.9 ± 5.1
700 99.1 ± 0.697.7 ± 1.194.0 ± 2.987.0 ± 5.0
750 99.4 ± 0.397.8 ± 1.495.2 ± 2.087.4 ± 5.7
800 99.6 ± 0.398.4 ± 1.296.7 ± 1.589.0 ± 5.3
850 99.7 ± 0.298.8 ± 0.997.5 ± 1.690.3 ± 5.3
900 99.8 ± 0.199.1 ± 0.597.8 ± 1.491.3 ± 4.4

Downloads