adhwiraj
/

ai-content-detector-roberta-v1

Text Classification

ai-content-detector

Eval Results (legacy)

text-embeddings-inference

Model card Files Files and versions

AI Content Detector - RoBERTa v1

Fine-tuned RoBERTa-base model for detecting AI-generated vs Human-written text.

Labels

Label ID	Label	Description
0	AI	AI-generated text
1	Human	Human-written text

Performance

Accuracy: 97.57%
F1 Score: 97.36%
Precision: 97.08%
Recall: 97.65%
ROC-AUC: 99.72%

Training Details

Base Model: roberta-base
Dataset: 3,706 samples (combined from 12 JSONL sources)
Epochs: 3 (best checkpoint at epoch 2)
Max Sequence Length: 512 tokens

Downloads last month: 12

Safetensors

Model size

0.1B params

Tensor type

F32

·

Model tree for adhwiraj/ai-content-detector-roberta-v1

Base model

FacebookAI/roberta-base

Finetuned

this model

Evaluation results

Test F1
self-reported

0.974
Test Accuracy
self-reported

0.976
Test ROC-AUC
self-reported

0.997