--- tags: Book --- # Round 11 Anti-Sybil Workflow ![](https://lucid.app/publicSegments/view/9a786de8-76f2-4c9d-b617-43e2757037c5/image.png) This is a TLDR for the rest of the Gitcoin Sybil Detection Book, laying out the general workflow and assignments for the Sybil detection microservices (blue boxes) and human evaluation (large orange box). Everything is described in greater detail in the relevant part of the book. * 8 September 2021 - Round 11 Begins * About 3-6 hours after beginning, manual invocation of the data collection, processing, ML microservices (BSCI) * First public report (notebook) of summary statistics produced (BSCI, automated) * First private report (csv) produced (BSCI, automated) * From then on, the microservices for data collection, processing, ML and public & private report run every 6 hours (BSCI, automated) * Roughly three (3) days after round 11 begins (13 September) human evaluation processes begin (Gitcoin & BSCI). Human evaluation entails: * `Prepare Evaluation` and `Prepare Human Evaluation Sheet` generated from microservices (BSCI) * Evaluation sheets assigned and distributed to human evaluators by Stream Cooordinator (GC) * Human Evaluators have three (3) days to complete their evaluations (GC) * Evaluations collected and confirmed by GC * Evaluations passed into `Retrieve Human Labels` microservice (BSCI) * Evaluations get human sanity/spot check for one (1) day to ensure no bad data enters the pipeline (BSCI) * Human evaluation process runs twice on that cycle * 3 days for data collection and ML processing * 3 days for human evaluation process * 1 day for human spot/sanity check * 24 September 2021 - Round Concludes * Subsequent final private report with round analysis.