naazimsnh02/TriageIQ-Qwen3-4B

TEXT GENERATIONConcurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jun 15, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

naazimsnh02/TriageIQ-Qwen3-4B is a 4 billion parameter Qwen3-based causal language model fine-tuned by naazimsnh02 for automated IT service-desk ticket triage. This model specializes in converting free-text IT support complaints into a single, structured, schema-valid JSON incident record. It is optimized for high accuracy in categorizing and assigning IT incidents, achieving 97.3% category accuracy and 96.0% assignment-group agreement on a held-out validation set.

Loading preview...

TriageIQ-Qwen3-4B: Specialized IT Incident Triage Model

TriageIQ-Qwen3-4B is a 4 billion parameter model, fine-tuned from Qwen/Qwen3-4B-Instruct-2507 by naazimsnh02, specifically designed for automated IT service-desk ticket triage. It transforms free-text IT support complaints into a structured JSON incident record, making it a fast-path component for predictive complaint-triage and SLA-aware routing engines.

Key Capabilities

  • Structured JSON Output: Converts IT complaints into a precise JSON object with fields like summary, category, urgency, impact, assignment_group, suggested_first_action, and confidence.
  • High Accuracy: Achieves 97.3% category accuracy and 96.0% assignment-group agreement on a held-out validation set, significantly outperforming the base model in these critical areas.
  • Schema Validity: Maintains a 100% schema-valid rate for its JSON outputs, ensuring reliable integration with downstream systems.
  • Optimized Training: Fine-tuned using bf16 LoRA (merged into base weights) on AMD Instinct MI300X hardware, ensuring efficient inference without PEFT requirements.

Good For

  • Automated IT Ticket Triage: Ideal for first-pass processing of IT service-desk tickets, structuring them for further automation.
  • Feeding Routing Engines: Provides a consistent JSON contract for deterministic routing and SLA engines.
  • Research & Development: Suitable for demo and research within the TriageIQ project, particularly for customer complaint classification and routing.

Limitations

  • Trained on synthetic, English-only, single-paragraph complaints; performance on other data types is unverified.
  • Limited to two assignment groups and four categories, coercing inputs into this taxonomy.
  • Not intended for safety-critical decisions or as the sole authority for ticket prioritization without a downstream deterministic engine.

Finetunes
1 models