geolip-bytelex / t1_matrix_ledger.json
AbstractPhil's picture
btx-e001 T-1: matrix size verdict — closed register saturates at 94 cells by 16KB; open-class digits grow 1276+ cells and flip modal shape to 2x2
f8b0b26 verified
Raw
History Blame Contribute Delete
1.96 kB
{
"32000": {
"bytes": 32138,
"tokens_A_flan": 9607,
"tokens_B_bert": 8271,
"tok_per_100b_A": 29.89,
"tok_per_100b_B": 25.74,
"co_boundary_rate": 0.923,
"matrix_nnz": 94,
"cells_at_90": 59,
"cells_at_99": 89,
"cells_at_999": 94,
"mass_total": 8148,
"frac_11_cells": 0.8085,
"frac_11_mass": 0.829,
"erank_11": 42.91,
"distinct_11_pairs": 76,
"shape_hist_top": {
"1x1": 76,
"2x1": 13,
"3x1": 4,
"1x2": 1
}
},
"64000": {
"bytes": 64234,
"tokens_A_flan": 19231,
"tokens_B_bert": 16552,
"tok_per_100b_A": 29.94,
"tok_per_100b_B": 25.77,
"co_boundary_rate": 0.9227,
"matrix_nnz": 94,
"cells_at_90": 59,
"cells_at_99": 89,
"cells_at_999": 94,
"mass_total": 16305,
"frac_11_cells": 0.8085,
"frac_11_mass": 0.8307,
"erank_11": 42.84,
"distinct_11_pairs": 76,
"shape_hist_top": {
"1x1": 76,
"2x1": 13,
"3x1": 4,
"1x2": 1
}
},
"128000": {
"bytes": 128208,
"tokens_A_flan": 38309,
"tokens_B_bert": 32916,
"tok_per_100b_A": 29.88,
"tok_per_100b_B": 25.67,
"co_boundary_rate": 0.9236,
"matrix_nnz": 94,
"cells_at_90": 60,
"cells_at_99": 90,
"cells_at_999": 94,
"mass_total": 32450,
"frac_11_cells": 0.8085,
"frac_11_mass": 0.8319,
"erank_11": 43.14,
"distinct_11_pairs": 76,
"shape_hist_top": {
"1x1": 76,
"2x1": 13,
"3x1": 4,
"1x2": 1
}
},
"256016": {
"bytes": 256016,
"tokens_A_flan": 79013,
"tokens_B_bert": 68746,
"tok_per_100b_A": 30.86,
"tok_per_100b_B": 26.85,
"co_boundary_rate": 0.8908,
"matrix_nnz": 1362,
"cells_at_90": 103,
"cells_at_99": 868,
"cells_at_999": 1297,
"mass_total": 65362,
"frac_11_cells": 0.13,
"frac_11_mass": 0.7983,
"erank_11": 53.7,
"distinct_11_pairs": 177,
"shape_hist_top": {
"2x2": 718,
"2x1": 379,
"1x1": 177,
"2x3": 59,
"1x2": 25,
"3x1": 4
}
}
}