ningpy/intent-detection-V2.0

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 12, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The ningpy/intent-detection-V2.0 model is a 7.6 billion parameter health-app user intent classifier and safety router, fine-tuned from Qwen2.5-7B-Instruct. It is specifically designed to categorize user messages into five states: PASS, REDFLAG, BLOCK, STEER, or COEXIST, with a focus on identifying emergency and personal-dose related queries. This model excels at safety classification, achieving a Safety F1 score of 92.3% and a Multi-label EM of 87.4%, making it highly effective for health-related conversational AI applications requiring robust safety and intent routing.

Loading preview...

Overview

ningpy/intent-detection-V2.0 is a specialized 7.6 billion parameter model, fine-tuned from Qwen2.5-7B-Instruct, functioning as a health-app user intent classifier and safety router. It processes user messages to determine their intent and potential safety implications, routing them into one of five distinct states: PASS, REDFLAG, BLOCK, STEER, or COEXIST (emergency + personal-dose).

Key Capabilities

  • Robust Safety Classification: Achieves a Safety F1 score of 92.3% (for REDFLAG \u222A BLOCK) and a REDFLAG F1 of 94.1%, indicating strong performance in identifying critical safety concerns.
  • Multi-label Intent Detection: Features a new multi-label capability with 76.2% coexist recall, allowing for more nuanced classification of complex user queries.
  • Structured Output: Provides a fixed JSON output schema with 6 keys, including intent, redflag, block, and entities (drugs, diseases, symptoms) extracted verbatim from user text.
  • Comprehensive Training: Trained on 54,404 samples across English, Chinese, and Bruneian Malay, with 57.1% of samples containing back-filled entities.

Use Cases

This model is ideal for health-related conversational AI systems that require precise intent detection and safety routing. It can be used to:

  • Automatically categorize user queries in health applications.
  • Identify and flag emergency situations or requests for personal medical advice.
  • Guide users towards appropriate resources or consultations based on their input.
  • Ensure safe and responsible interactions within health-focused platforms by routing messages to specific actions (e.g., emergency guidance, refusing dose advice, referring to a doctor).

Finetunes
1 models