kelsbeans/qwen3-1.7b-digestive-coach-n781

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 20, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The kelsbeans/qwen3-1.7b-digestive-coach-n781 is a 1.7 billion parameter Qwen3-based language model, fine-tuned by kelsbeans using QLoRA on a specialized digestive health dataset. This model is specifically optimized to act as a digestive wellness coach, demonstrating significant improvements in parse rate, schema validity, and adherence to conversational specifications compared to its base model. It excels in generating structured, relevant responses within the domain of digestive health, making it suitable for applications requiring specialized, non-medical advice in this area.

Loading preview...

Digestive Wellness Coach Model

This model, kelsbeans/qwen3-1.7b-digestive-coach-n781, is a specialized fine-tune of the Qwen/Qwen3-1.7B base model. Developed by kelsbeans, it leverages Unsloth QLoRA for efficient adaptation. The training utilized 781 rows from the kelsbeans/digestive-coach-dataset, focusing on digestive health scenarios.

Key Capabilities and Performance

This fine-tuned model demonstrates substantial improvements over its base model in domain-specific tasks:

  • Enhanced Output Structure: Achieves a 100.0% parse rate and 100.0% schema validity on held-out evaluation scenarios, a significant leap from the base model's 12.5% and 2.5% respectively.
  • Improved Spec Adherence: Shows a 20.0% spec adherence, outperforming the base model (0.0%) and even frontier prompted cells (15%).
  • Higher Field Accuracy: Reaches 39.7% field accuracy compared to the base's 26.7%.
  • Qualitative Improvements: Judge evaluations (using Anthropic's Claude Sonnet 5) indicate superior integrated voice (4.78 vs 2.88), honest uncertainty (4.09 vs 2.19), and no manufactured agreement (4.47 vs 2.62).

Use Cases

This model is designed as a research artifact for fine-tuning projects focused on data-to-behavior transformation. It is particularly well-suited for:

  • Developing AI assistants that provide non-medical, informational guidance on digestive wellness.
  • Applications requiring structured and consistent output in health-related conversational agents.
  • Exploring specialized domain adaptation for smaller language models.

Important Note: This model is a research artifact and not intended to provide medical advice.

Finetunes
2 models