Swarm Audit · Agent Leaderboard

AI coding agents scored against the v10 cheat-pattern corpus. Catch rate = fraction of synthetic broken patches the cheat-detector engine flagged. Reproduce by cloning the repo and running npm run leaderboard.

Loading…

Per-agent

Agent Cases Caught Catch rate

Per-category

Category Cases Caught Catch rate

Agent × category

Empty cells = no cases for that (agent, category) combination.