MadeAgents/Hammer2.0-7b: Robust Function Calling for On-Device Agents
MadeAgents/Hammer2.0-7b is a 7.6 billion parameter language model developed by MadeAgents, specifically engineered for advanced function calling capabilities. It is fine-tuned from the Qwen 2.5 series and Qwen 2.5 Coder series, utilizing innovative function masking techniques to enhance its ability to accurately interpret and execute tool calls.
Key Capabilities & Features
- Exceptional Function Calling: Achieves strong performance on the Berkeley Function-Calling Leaderboard (BFCL-v3), often outperforming larger models in its class.
- Specialized Training: Trained on a combination of the APIGen Function Calling Datasets (60,000 samples) and the xlam-irrelevance-7.5k dataset.
- Robust Generalization: Demonstrates stable performance across various academic benchmarks, indicating strong generalization beyond specific function calling tasks.
- Lightweight & Efficient: Part of a series of lightweight models designed for on-device agentic applications.
Ideal Use Cases
- On-Device Agentic Applications: Perfect for developing intelligent agents that require precise tool use without heavy computational overhead.
- Automated Workflows: Can be integrated into systems needing reliable function execution based on natural language prompts.
- Tool-Augmented LLMs: Enhances the capabilities of language models by enabling them to interact with external tools and APIs effectively.
For more technical details, refer to the research paper Hammer: Robust Function-Calling for On-Device Language Models via Function Masking and the Hammer GitHub repository.