HC3 Human vs ChatGPT Detector
This model was fine-tuned on the HC3 dataset to classify whether an answer was written by a human or generated by ChatGPT.
Model
Base model: sentence-transformers/all-MiniLM-L6-v2
Task:
- Label 0: Human
- Label 1: ChatGPT
Results
- Baseline accuracy (frozen sentence embeddings + logistic regression): 84.51%
- Fine-tuned test accuracy: 98.93%
Training
- Learning rate:
2e-5 - Epochs:
5 - Batch size:
32 - Maximum sequence length:
256
Notes
The fine-tuned model achieved substantially higher test accuracy than the frozen-embedding logistic regression baseline. However, the model was trained and evaluated on HC3, so its performance may not generalize to AI-generated text from other datasets or newer language models.
- Downloads last month
- 17