TSUM CryptoBERT Sentiment Classifier
암호화폐 뉴스 헤드라인 → bearish / neutral / bullish 분류 모델.
학습 결과
- Accuracy: 0.6659
- Macro F1: 0.6447
사용법
from transformers import pipeline
clf = pipeline("text-classification", model="space1637/tsum-cryptobert-sentiment")
clf("Bitcoin ETF inflows surge as institutional adoption accelerates")
학습 데이터
- StephanAkkerman/financial-tweets-crypto
- SahandNZ/cryptonews-with-price-momentum-labels
베이스 모델
- ElKulako/cryptobert + LoRA (r=16, alpha=32)