"The highlighted tokens are primarily Japanese content words, such as nouns, verbs, and adjectives, as well as some grammatical particles and endings that are important for sentence structure and meaning. There is also a notable presence of corrupted or placeholder tokens, likely representing unreadable or missing characters. The activations focus on semantically significant morphemes and key syntactic elements within Japanese sentences."
Score Type | Accuracy | Precision | Recall | F1 score | TPR | TNR | FPR | FNR |
---|---|---|---|---|---|---|---|---|
detection | 0.68 | 1.0 | 0.36 | 0.529 | 0.36 | 1.0 | 0.0 | 0.64 |
fuzz | 0.64 | 0.769 | 0.4 | 0.526 | 0.4 | 0.88 | 0.12 | 0.6 |