"The highlighted tokens are primarily Korean morphemes, words, or short phrases that serve as grammatical particles, noun or verb endings, or key content words within sentences. These tokens often mark syntactic boundaries, indicate relationships between sentence elements, or represent important semantic units such as subjects, objects, or actions. The activations tend to focus on tokens that are functionally or semantically significant for understanding sentence structure and meaning in Korean text."
Score Type | Accuracy | Precision | Recall | F1 score | TPR | TNR | FPR | FNR |
---|---|---|---|---|---|---|---|---|
detection | 0.64 | 1.0 | 0.28 | 0.438 | 0.28 | 1.0 | 0.0 | 0.72 |
fuzz | 0.69 | 1.0 | 0.38 | 0.551 | 0.38 | 1.0 | 0.0 | 0.62 |