imagining syntax experiment results

← depth

n-layer_3

done · peak unseen_mismatch 92.1

Configuration

n_layer
3
n_head
4
n_embd
256
base_seed
42

Figures

accuracy_vs_alpha.png
accuracy_vs_alpha.png

Results

paramseen_matchseen_mismatchunseen_matchunseen_mismatch
0.0 100.0 ± 0.029.9 ± 19.7100.0 ± 0.030.2 ± 19.3
0.1 100.0 ± 0.016.6 ± 20.2100.0 ± 0.016.8 ± 20.1
0.2 100.0 ± 0.023.8 ± 28.4100.0 ± 0.023.2 ± 27.5
0.3 97.0 ± 9.05.6 ± 10.896.6 ± 10.25.4 ± 10.8
0.4 100.0 ± 0.020.9 ± 31.2100.0 ± 0.021.2 ± 31.5
0.5 100.0 ± 0.019.2 ± 28.5100.0 ± 0.019.0 ± 27.3
0.6 100.0 ± 0.028.8 ± 31.2100.0 ± 0.029.3 ± 31.0
0.7 100.0 ± 0.012.3 ± 16.8100.0 ± 0.012.6 ± 17.0
0.8 100.0 ± 0.018.8 ± 18.1100.0 ± 0.019.0 ± 18.1
0.9 100.0 ± 0.013.0 ± 17.5100.0 ± 0.013.2 ± 17.9
1.0 100.0 ± 0.040.8 ± 35.0100.0 ± 0.040.8 ± 35.1
1.1 100.0 ± 0.035.0 ± 40.4100.0 ± 0.035.0 ± 40.4
1.2 100.0 ± 0.054.0 ± 36.6100.0 ± 0.054.0 ± 36.5
1.3 100.0 ± 0.078.8 ± 18.599.9 ± 0.478.9 ± 17.8
1.4 99.9 ± 0.285.9 ± 14.199.3 ± 1.683.4 ± 14.6
1.5 99.8 ± 0.294.9 ± 7.999.0 ± 1.490.8 ± 9.3
1.6 99.5 ± 0.593.4 ± 11.998.4 ± 1.591.3 ± 11.3
1.7 99.4 ± 0.498.3 ± 1.196.9 ± 1.892.1 ± 2.9
1.8 99.5 ± 0.495.8 ± 4.694.9 ± 3.384.9 ± 6.2
1.9 99.1 ± 0.398.0 ± 0.892.1 ± 2.685.4 ± 4.9
2.0 99.1 ± 0.297.6 ± 1.789.6 ± 2.583.2 ± 4.9
2.1 99.3 ± 0.397.6 ± 1.589.9 ± 4.077.4 ± 11.4
2.2 99.3 ± 0.398.5 ± 0.886.1 ± 3.479.2 ± 3.8
2.3 99.2 ± 0.597.2 ± 2.685.2 ± 3.072.3 ± 10.3
2.4 99.2 ± 0.397.8 ± 1.582.8 ± 2.573.5 ± 6.8
2.5 99.4 ± 0.398.5 ± 0.880.8 ± 2.871.6 ± 4.6
2.6 99.4 ± 0.298.2 ± 1.081.5 ± 4.373.8 ± 4.9
2.7 99.6 ± 0.299.1 ± 1.080.6 ± 2.072.6 ± 5.9
2.8 99.4 ± 0.399.2 ± 0.476.0 ± 5.371.2 ± 6.6
2.9 99.6 ± 0.298.3 ± 2.078.7 ± 2.967.0 ± 6.3
3.0 99.7 ± 0.298.8 ± 0.975.0 ± 3.364.7 ± 6.1

Downloads