List your company

Intrinsic Tokenization on FLORES+ (test)

69.2Vocabulary UtilisationBias
2.6419.9237.254.48May 21, 2026

Evaluation results

MethodLinks
BiasVocabulary Size=8k2026.0569.20.00936.794.29156.71.96121.90.0082
BPEVocabulary Size=8k2026.0567.70.0096.754.34131.11.91123.50.0081
DetVocabulary Size=8k2026.0563.80.00856.734.31127.81.93123.40.0081
IntVocabulary Size=8k2026.05620.00516.414.5457.61.52136.70.0073
BiasVocabulary Size=16k2026.0548.10.01547.374.84233.92.19102.90.0097
BPEVocabulary Size=16k2026.0546.50.01427.254.9192.22.11107.90.0093
DetVocabulary Size=16k2026.0544.30.01387.294.86196.32.16105.80.0095
IntVocabulary Size=16k2026.0542.70.00846.764.889.11.77124.40.008
BiasVocabulary Size=32k2026.0529.40.02167.855.27322.52.3989.40.0112
BPEVocabulary Size=32k2026.0528.70.02017.715.27272.52.393.50.0107
DetVocabulary Size=32k2026.0527.70.027.765.25285.42.36910.011
IntVocabulary Size=32k2026.0526.60.01327.225.13153.12.06108.80.0092
BiasVocabulary Size=64k2026.0517.50.03348.916.37519.62.6569.10.0145
BPEVocabulary Size=64k2026.0516.70.03018.796.68432.42.5772.70.0138
DetVocabulary Size=64k2026.0516.60.03138.846.33475.62.6369.90.0143
IntVocabulary Size=64k2026.0515.80.01957.825.55245.62.390.10.0111
BiasVocabulary Size=128k2026.05100.04599.746.85779.82.9257.60.0174
DetVocabulary Size=128k2026.059.70.04429.76.83742.12.9157.90.0173
BPEVocabulary Size=128k2026.059.40.04059.527.25632.32.83610.0164
IntVocabulary Size=128k2026.059.30.03218.726.37469.72.6669.60.0144
BiasVocabulary Size=256k2026.055.90.062410.467.61,140.23.1949.90.02
DetVocabulary Size=256k2026.055.80.060810.417.561,1073.1949.80.0201
BPEVocabulary Size=256k2026.055.40.054810.287.719513.151.90.0193
IntVocabulary Size=256k2026.055.20.03919.096.74590.32.863.90.0157