mcphorizon/mcp-horizon-support-v1
mcphorizon/mcp-horizon-support-v1 is a 4 billion parameter language model, fine-tuned from Qwen3-4B, specifically designed for question-answering related to MCP Horizon product documentation. It leverages QLoRA for efficient fine-tuning and is optimized to provide support and information on MCP Horizon topics. This model is ideal for integrating into customer support systems or knowledge bases focused on MCP Horizon products.
Loading preview...
Model Overview
mcphorizon/mcp-horizon-support-v1 is a 4 billion parameter language model, fine-tuned from the Qwen3-4B base model. Its primary purpose is to provide accurate question-answering and support for topics related to MCP Horizon product documentation. The model was fine-tuned using QLoRA (4-bit) via Unsloth, making it efficient for deployment and inference.
Key Capabilities
- Specialized Knowledge: Fine-tuned on specific MCP Horizon product documentation, enabling it to answer detailed questions about the product.
- Efficient Fine-tuning: Utilizes QLoRA with a rank of 16 and alpha of 32, trained over 5 epochs with 534 examples.
- Accessibility: A GGUF Q4_K_M version is available for use with Ollama/llama.cpp, facilitating local deployment.
Performance Insights
Evaluated on 60 examples using DeepEval, the model achieved an Answer Relevancy score of 0.86 with an 80% pass rate. While its Correctness (GEval) and Completeness (GEval) scores were 0.32 (22% pass rate) and 0.35 (25% pass rate) respectively, its strength lies in providing relevant answers based on its specialized training data.
Good For
- Customer Support: Automating responses to common queries about MCP Horizon products.
- Internal Knowledge Bases: Providing quick access to information for support agents or internal teams.
- Documentation Q&A: Building interactive Q&A systems directly from product manuals and guides.