videogameaetoros/Ai_detector

TutorCat AI Detector (ModernBERT)

Binary human vs AI text classifier for TutorCat integrity checks.

• Base: answerdotai/ModernBERT-base
• Labels: 0 = human, 1 = ai
• Train data: integrity_human_ai_surface_v2_no_train.jsonl (~35k rows)
• External holdout: accuracy 99.5%, AUROC 0.9999 (200 essays)

Load

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

repo = "videogameaetoros/Ai_detector"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForSequenceClassification.from_pretrained(repo)
model.eval()

text = "Your essay here..."
enc = tok(text, return_tensors="pt", truncation=True, max_length=384)
with torch.no_grad():
    probs = torch.softmax(model(**enc).logits, dim=-1)[0]
print({"human": float(probs[0]), "ai": float(probs[1])})

Space

Used by https://huggingface.co/spaces/videogameaetoros/ai_detector via INTEGRITY_MODEL_REPO=videogameaetoros/Ai_detector.