Overview

nopenet/nope-edge is a 4 billion parameter model developed by NopeNet, specifically fine-tuned for crisis classification in text. It identifies safety-critical content such as suicidal ideation, self-harm, abuse, and violence, providing detailed chain-of-thought reasoning for its classifications. The model outputs structured XML, including a reflection on its decision and specific risk attributes like subject, type, severity, and imminence.

Key Capabilities

• Crisis Detection: Accurately identifies various crisis types, including suicide, self_harm, violence, abuse, and exploitation.
• Chain-of-Thought Reasoning: Provides a <reflection> explaining the classification, detailing detected signals and contextual factors.
• Structured Output: Classifies risks into XML format with attributes like subject (self/other), type, severity (mild to critical), and imminence (chronic to emergency).
• Contextual Understanding: Trained to differentiate genuine crisis signals from hyperbolic language or recovery narratives.

Good For

• Content Moderation: Automatically flagging safety-critical content in user-generated text.
• Mental Health Support: Identifying users at risk in online communities or support platforms.
• Research & Evaluation: Academic and nonprofit studies on crisis detection and AI safety.

Important Limitations

It's crucial to note that this model provides probabilistic signals, not clinical assessments. It is not a medical device and should never be the sole basis for intervention decisions. Human review is always recommended for flagged content.