SiBot: A hybrid framework for fine-grained semantic attribution of AI-generated text and its production deployment
Md Sadiq Iqbal, Mohammod Abul Kashem · 2026
Read SiBot: A hybrid framework for fine-grained semantic attribution of AI-generated text and its production deployment by Md Sadiq Iqbal, Mohammod Abul Kashem (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0357405.txt.
Description
Md Sadiq Iqbal and Mohammod Abul Kashem published a 2026 paper in PLOS ONE arguing that asking whether a text is human or machine-written is the wrong question. Forensics wants to know which model wrote it; universities want to know whether a human collaborated. Their system, SiBot, treats detection as source attribution across many classes. They built and released two benchmark corpora. One spans more than twenty AI sources; another models pure and human-AI hybrid writing.
On the harder many-source corpus, the best classical combination reached only middling accuracy, showing how tough attribution is. On the ten-class corpus, their hybrid design, blending word-frequency features with a transformer's contextual embeddings, scored above ninety percent, beating the strongest transformer alone by a wide margin. The framework is publicly deployed. Detectors, they say, must move from binary verdicts to graded, source-aware answers among hybrid drafts.
Catalog record
| CREATOR | Md Sadiq Iqbal, Mohammod Abul Kashem |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 116553 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |