Who Wrote This Line? Evaluating the Detection of LLM-Generated Classical Chinese Poetry

This project is a classic academic reference implementation for a niche AI detection task. While it addresses a culturally significant domain (Classical Chinese poetry), its technical moat is virtually non-existent. Detection of AI-generated text is a 'cat-and-mouse' game where the generative models (the 'cats') generally hold the advantage; as LLMs like Baidu's Ernie or specialized models like Jiushige improve their grasp of classical prosody, the statistical signatures this project likely relies on will disappear. The quantitative signals (0 stars but 8 forks in 6 days) suggest an academic context where peers or students are forking the code for replication, but it lacks broader market traction. From a competitive standpoint, any frontier lab or major Chinese tech platform (Tencent, Alibaba) could integrate similar detection or, more likely, watermarking directly into their models, making third-party detection tools obsolete. The defensibility is low because the underlying classifier techniques (likely BERT-based or statistical) are commodity patterns applied to a specific dataset.