RADAR detector | trigger=best | AUROC=0.3406

README.md

@@ -0,0 +1,47 @@
+---
+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
+- **Best macro AUROC**: 0.3406
+- **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-Detector-PANCLEF")
+model     = RobertaForSequenceClassification.from_pretrained("Shushant/ADAL-Detector-PANCLEF")
+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

config.json

@@ -0,0 +1,28 @@
+{
+  "add_cross_attention": false,
+  "architectures": [
+    "RobertaForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "is_decoder": false,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 514,
+  "model_type": "roberta",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "pad_token_id": 1,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.3.0",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 50265
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff550cff2438cc62fe0fc8c31f7c12a483cd17c512db0d862b55ee16f6870c96
+size 1421495360

tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": "<s>",
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "errors": "replace",
+  "is_local": false,
+  "mask_token": "<mask>",
+  "model_max_length": 512,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "tokenizer_class": "RobertaTokenizer",
+  "trim_offsets": true,
+  "unk_token": "<unk>"
+}