HW1 HC3 Detector
Fine-tuned classifier for detecting human vs. ChatGPT text (HC3 dataset).
Labels
LABEL_0= humanLABEL_1= ChatGPT
Results
| Model | Test accuracy |
|---|---|
| Baseline (frozen embeddings + logistic regression) | 84.49% |
| Fine-tuned (full transformer fine-tuning) | 99.51% |
Training Parameters
| Parameter | Value |
|---|---|
| Base model | sentence-transformers/all-MiniLM-L6-v2 |
| Optimizer | AdamW |
| Learning rate | 2e-5 |
| Epochs | 5 |
| Batch size | 32 |
| Max sequence length | 256 tokens |
| Seed | 42 |
Usage
from transformers import pipeline
clf = pipeline("text-classification", model="Dhanushya2000/hw1-hc3-detector")
- Downloads last month
- 30
Model tree for Dhanushya2000/hw1-hc3-detector
Base model
nreimers/MiniLM-L6-H384-uncased Quantized
sentence-transformers/all-MiniLM-L6-v2