cs-552-2026-llmfao/group_model
The cs-552-2026-llmfao/group_model is a merged language model based on Qwen/Qwen3-1.7B, created using the DARE TIES method. This 1.7 billion parameter model integrates specialized components for general language understanding, mathematical reasoning, safety, and multilingual capabilities. It is designed to offer a balanced performance across diverse tasks by combining strengths from multiple fine-tuned Qwen3-1.7B variants.
Loading preview...
Model Overview
The cs-552-2026-llmfao/group_model is a 1.7 billion parameter language model derived from the Qwen/Qwen3-1.7B base model. It was constructed using the DARE TIES merge method, which combines multiple specialized models into a single, more versatile model.
Key Capabilities
This model integrates expertise from several distinct Qwen3-1.7B variants, aiming for balanced performance across:
- General Language Understanding: Incorporates a component focused on broad linguistic tasks.
- Mathematical Reasoning: Includes a specialized component for handling mathematical problems.
- Safety: Features a component designed to enhance model safety and mitigate harmful outputs.
- Multilingual Support: Integrates a component to improve performance across multiple languages.
Merge Details
The merge process utilized mergekit and combined the base Qwen/Qwen3-1.7B with fine-tuned versions for general, safety, math, and multilingual applications. Each component was assigned specific density and weight parameters during the DARE TIES merge, with the base model serving as the foundation. The final model uses bfloat16 for its data type and includes int8_mask for parameters.