Views
No views yet
| Metric | Score |
|---|---|
| Sentence Accuracy | 95.6% |
| Macro F1 Score | 95.6% |
| Word Accuracy | 95.5% |
gpt2: GPT-2 generatedgptneo: GPT-Neo generatedgptj: GPT-J generatedllama: LLaMA generatedgpt3re: GPT-3 rewrittenhuman: Human written1import torch
2from model import ModelWiseTransformerClassifier
3
4# Load model
5model = torch.load("seqxgpt_transformer.pt")
6model.eval()
7
8# Inference requires feature extraction from 4 LLMs (GPT-2, GPT-Neo, GPT-J, LLaMA)
9# See the original SeqXGPT repo for full pipeline1@misc{wang2023seqxgpt,
2 title={SeqXGPT: Sentence-Level AI-Generated Text Detection},
3 author={Pengyu Wang and Linyang Li and Ke Ren and Botian Jiang and Dong Zhang and Xipeng Qiu},
4 year={2023},
5 eprint={2310.08903},
6 archivePrefix={arXiv},
7 primaryClass={cs.CL}
8}