CircleGuardBench: New Standard for Evaluating AI Moderation Models
• 59
AI & ML interests
Runtime AI safety and alignment infrastructure
Recent Activity
View all activity
Articles
Team members 11 private
Organization Card
White Circle is an AI control layer for businesses.
- Automatically test AI systems for failures, jailbreaks, and hallucinations
- Enforce guardrails and custom policies in real time
- Monitor behavior with analytics, metrics, and user insights
- Optimize performance with prompt engineering and model routing
Learn more about us at whitecircle.ai
models 0
None public yet
datasets 5
whitecircle/math-reasoning-sft
Viewer • Updated • 96
whitecircle/swe-rebench-v2-glm-5.1-pi-agent-successful-traces
Viewer • Updated • 7.78k • 76 • 2
whitecircle/swe-rebench-v2-clean-python-tasks
Viewer • Updated • 4.34k • 95 • 1
whitecircle/killbench
Viewer • Updated • 1.37M • 72 • 6
whitecircle/circleguardbench_public
Viewer • Updated • 980 • 50 • 2