File size: 1,521 Bytes
a252672 b9e7c1a a252672 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 | ---
language: en
license: apache-2.0
tags:
- text-classification
- ai-generated-text-detection
- roberta
- adversarial-training
metrics:
- roc_auc
---
# RADAR Detector (RoBERTa-large)
Adversarially trained AI-generated text detector based on the RADAR framework
([Hu et al., NeurIPS 2023](https://arxiv.org/abs/2307.03838)), extended with
a multi-evasion attack pool for robust detection.
## Training
- **Base model**: `roberta-large`
- **Dataset**: [RAID](https://huggingface.co/datasets/liamdugan/raid) (Dugan et al., ACL 2024)
- **Evasion attacks seen during training**: t5_paraphrase, synonym_replacement, homoglyphs, article_deletion, misspelling, number_swap, whitespace_addition, upper_lower_swap, zero_width_space, insert_paragraphs, alternative_spelling
- **Best macro AUROC**: 0.9782
- **Generators**: chatgpt, gpt2, gpt3, gpt4, cohere, cohere-chat, llama-chat,
mistral, mistral-chat, mpt, mpt-chat
## Usage
```python
from transformers import RobertaTokenizer, RobertaForSequenceClassification
import torch
tokenizer = RobertaTokenizer.from_pretrained("Shushant/adal_v5_raid")
model = RobertaForSequenceClassification.from_pretrained("Shushant/adal_v5_raid")
model.eval()
text = "Your text here."
enc = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
with torch.no_grad():
probs = torch.softmax(model(**enc).logits, dim=-1)[0]
print(f"P(human)={probs[1]:.3f} P(AI)={probs[0]:.3f}")
```
## Label mapping
- Index 0 → AI-generated
- Index 1 → Human-written
|