FlexGuard: Continuous Risk Scoring for Strictness-Adaptive LLM Content Moderation

Researchers introduce FlexGuard, a new AI content moderation system that provides continuous risk scoring instead of binary decisions, allowing platforms to adapt moderation strictness as needed. The system addresses limitations of existing guardrail models that break down when content moderation requirements change across platforms or over time.

• →FlexGuard outputs calibrated continuous risk scores rather than binary safe/unsafe classifications for LLM content moderation.
• →Existing moderation models show substantial performance degradation when strictness requirements change across platforms.
• →FlexBench benchmark enables controlled evaluation of moderators under multiple strictness regimes.
• →The system uses risk-alignment optimization and threshold selection strategies to adapt to different strictness levels.
• →FlexGuard demonstrates higher accuracy and improved robustness compared to existing binary moderation approaches.