NotoriousH2/Qwen3-0.6B-JSON-SFT

TEXT GENERATIONConcurrent Unit Cost:1Model Size:0.8BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 19, 2026Architecture:Transformer Featherless Exclusive Cold

NotoriousH2/Qwen3-0.6B-JSON-SFT is a 0.8 billion parameter model based on the Qwen3-0.6B architecture, fine-tuned for extracting meeting minutes into JSON format. Developed by NotoriousH2 as part of an SLLM practice curriculum improvement experiment, it demonstrates high parsing and schema compliance rates for structured data extraction. This model is specifically optimized for converting unstructured meeting notes into a structured JSON output.

Loading preview...

Model Overview

NotoriousH2/Qwen3-0.6B-JSON-SFT is a specialized 0.8 billion parameter language model derived from the Qwen3-0.6B architecture. It was developed by NotoriousH2 as an experimental output for an SLLM practice curriculum improvement project, focusing on enhancing structured data extraction capabilities.

Key Capabilities

This model is specifically fine-tuned for a unique task: converting unstructured meeting minutes into a structured JSON format. Its training involved 1,199 examples from the meeting_to_json_train.jsonl dataset, utilizing a full fine-tuning approach without a 'no-think' strategy.

Performance Metrics

Evaluated on a 300-sample dataset, the model achieved notable performance in JSON extraction:

  • Parse Rate: 98.0%
  • Schema Compliance: 96.3%
  • Semantics Pass Rate: 96.1%
  • Field F1 Mean: 51.4% (a significant improvement over the base model's 41.6%)

Use Cases

This model is ideal for applications requiring precise and automated extraction of key information from meeting transcripts or notes into a machine-readable JSON structure. Its high parse and schema compliance rates make it suitable for integrating meeting data into databases, analytics platforms, or other structured data systems.

Finetunes
2 models