Moderation thresholds

Flag threshold: 0.35 Remove threshold: 0.70

Classifier tuning

Vocabulary sensitivity: 1.00× Stream speed: 1.2 posts/s

Scenario mix

Current post

Tokens scored—

Weighted sum—

Toxicity score—

Decision—

Session stats

Posts scored0

Approved0

Flagged0

Removed0

Approval rate—

Each post is tokenized on whitespace/punctuation, every known token contributes its learned toxicity weight (scaled by sensitivity) to a running sum, and a logistic sigmoid turns that sum into a 0–1 toxicity score. The score is compared against the two thresholds to sort the post into approved / flagged / removed. The top panel highlights exactly which tokens pushed the score up (red) or down (green).