Module E: Ensemble
Run one piece of text through the statistical detector, the trained classifier and the retrieval ledger at once. Each verdict stays visible, and the combined estimate abstains when they contradict each other.
How it works and its limits
The combined estimate pools only B and C, the two modules that produce probabilities, using inverse-variance weighting. F is shown alongside because it answers a different question: has this suite seen this text before?
Watermarking (A) needs a known key and file provenance (D) checks signatures rather than style, so neither fits this kind of combination. Read the design notes.
or try
Results appear here: a combined estimate, each module’s own verdict, and a sentence-by-sentence heatmap.