Fine-tuned GPT-3.5 and prompt-guided GPT-4.5 can classify multiple suicide-related factors in psychiatric notes with up to 0.94 partial-match accuracy, though rare labels remain difficult.
We extend the classical confusion matrix to the power- set setting, enabling granular inspection of hallucination (false-positive) versus omission (false- negative) patterns
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Multi-Label Classification with Generative AI Models in Healthcare: A Case Study of Suicidality and Risk Factors
Fine-tuned GPT-3.5 and prompt-guided GPT-4.5 can classify multiple suicide-related factors in psychiatric notes with up to 0.94 partial-match accuracy, though rare labels remain difficult.