SeongryongJung/qwen3-4b-tooluse-rlsd-ema005

TEXT GENERATIONConcurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 2, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The SeongryongJung/qwen3-4b-tooluse-rlsd-ema005 is a 4 billion parameter language model fine-tuned from Qwen/Qwen3-4B. It specializes in tool use, having been optimized with RLSD (EMA 0.05) on a dedicated tooluse dataset. This model demonstrates a validation performance of 60.11% mean@16 on tool use tasks, making it suitable for applications requiring external tool interaction.

Loading preview...

Model Overview

This model, qwen3-4b-tooluse-rlsd-ema005, is a 4 billion parameter language model derived from the Qwen/Qwen3-4B architecture. Its primary distinction lies in its fine-tuning process, which utilized RLSD (Reinforcement Learning from Simulated Dialog) with an Exponential Moving Average (EMA) of 0.05, specifically targeting the tooluse split of a dataset.

Key Capabilities

  • Tool Use Optimization: The model is explicitly fine-tuned for scenarios requiring interaction with external tools, indicating a strong capability in interpreting and executing tool-related instructions.
  • Performance Metrics: Achieved a final validation performance of 60.11% (mean@16) on tool use tasks, with a peak performance of 63.42% during training.
  • Context Length: Inherits a substantial context length of 32768 tokens, allowing for complex tool-use scenarios and longer interaction histories.

Good For

  • Agentic Applications: Ideal for developing AI agents that need to leverage external APIs or tools to complete tasks.
  • Automated Workflows: Suitable for integrating into systems where the model needs to call specific functions or interact with software components.
  • Research in RLSD: Provides a practical example of a model fine-tuned with RLSD for tool-use capabilities.

Finetunes
1 models