SKMT

Observations

• MorphBPE/SKMT-Style Tokenizer Smoke This is a constrained-BPE pipeline check on the current tiny Polish gold set, not a final tokenizer result. system | cases | boundary F1 | fertility | align P | align R standard_bpe_smoke | 12 | 0.510 | 3.000 | 0.544 | 0.597 morph_bpe_constrained_smoke | 12 | 0.925 | 3.167 | 0.889 | 1.000 ↗

Related

mentions: SOTA 2026 Baseline Matrix, MorphBPE/SKMT-Style Tokenizer Smoke, Polish Morpheme Boundary Benchmark Protocol
defined-by: MorphBPE/SKMT-Style Tokenizer Smoke

Local graph

Provenance