This model scores (i.e., outputs continuous values) that quantify the hatefulness of input text (typically social media comments). It is built on the work of Kennedy et al. that developed a measurement scale for hate speech.

Usage

python

from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("ucberkeley-dlab/mhs-scorer-modernbert-large", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained("AnswerDotAI/ModernBERT-large")

inputs = tokenizer("your text here", return_tensors="pt")
scores = model(**inputs)