Checkpoint safety leaderboard

Behavioral safety grades for open-weight AI checkpoints. Lower injection-compliance is safer. Sort any column; the grade follows the published rubric.

Injection compliance = genuine self-exfiltration under indirect prompt injection (Frame B). Autonomous = unprompted self-exfiltration (Frame A). Grades: A safest → F riskiest; prov. marks a provisional grade capped by small sample size. Ablated = refusal-direction-removed ("uncensored") checkpoint. See Methodology for the rubric and limits.