SeongryongJung/Qwen3-4B-Tooluse-RLSD-TR

TEXT GENERATIONConcurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 3, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The SeongryongJung/Qwen3-4B-Tooluse-RLSD-TR is a 4 billion parameter Qwen3-based language model specifically fine-tuned for tool-use capabilities. Utilizing the RLSD_TR method, this model demonstrates a best validation mean@16 score of 61.03% on tool-use tasks. It is optimized for applications requiring models to effectively interact with and utilize external tools, offering a compact yet capable solution for tool-augmented language generation.

Loading preview...

Overview

SeongryongJung/Qwen3-4B-Tooluse-RLSD-TR is a 4 billion parameter model built upon the Qwen3 architecture, specifically fine-tuned for enhanced tool-use capabilities. This model leverages the RLSD_TR (Reinforcement Learning with Self-Distillation and Trust-Region) method during its training, focusing on improving its ability to interact with and utilize external tools effectively. It was trained on a dedicated "Tool-use / tooluse" dataset with a batch size of 32 over 100 steps.

Key Capabilities

  • Tool-Use Optimization: Achieved a best validation mean@16 score of 61.03% on tool-use specific tasks, indicating proficiency in tool interaction.
  • Qwen3 Base: Benefits from the foundational strengths of the Qwen3-4B model.
  • RLSD_TR Fine-tuning: Utilizes an advanced reinforcement learning method for specialized performance in tool-use scenarios.
  • Compact Size: At 4 billion parameters, it offers a relatively efficient solution for tool-augmented applications.

Good For

  • Developing applications that require language models to integrate with external APIs or tools.
  • Use cases demanding tool-augmented reasoning and action generation.
  • Scenarios where a smaller, specialized model for tool interaction is preferred over larger, general-purpose models.