SeongryongJung/Qwen3-4B-Tooluse-GRPO-TR
The SeongryongJung/Qwen3-4B-Tooluse-GRPO-TR is a 4 billion parameter Qwen3-based causal language model, fine-tuned by SeongryongJung using the GRPO method. Optimized specifically for tool-use capabilities, this model achieves a best validation mean@16 score of 60.29% on tool-use datasets. It is designed for applications requiring effective function calling and interaction with external tools, leveraging a 32768 token context length.
Loading preview...
Overview
This model, SeongryongJung/Qwen3-4B-Tooluse-GRPO-TR, is a 4 billion parameter variant of the Qwen3 architecture, specifically fine-tuned for tool-use applications. Developed by SeongryongJung, it utilizes the GRPO (Generalized Reinforcement Learning with Policy Optimization) method for training, focusing on enhancing its ability to interact with external tools and APIs.
Key Capabilities
- Tool-use Optimization: The model is explicitly trained on tool-use datasets, demonstrating a best validation
mean@16score of 60.29%. - GRPO Fine-tuning: Leverages the GRPO training method, indicating a specialized approach to reinforcement learning for improved performance in its target domain.
- Qwen3 Base: Built upon the Qwen3-4B base model, providing a strong foundation for language understanding and generation.
- Extended Context Window: Supports a maximum model length of 10240 tokens, with a max response length of 8192 tokens, beneficial for complex tool-use scenarios.
Good For
- Function Calling: Ideal for applications requiring the model to accurately identify and call external functions or tools based on user prompts.
- Agentic Workflows: Suitable for building AI agents that need to interact with various systems and services.
- Automated Task Execution: Can be integrated into systems that automate tasks by leveraging external APIs and tools.