Qwen3_0.6B_Financial_Tool_Calling_LoRA

Qwen3 0.6B Financial Fundamentals Tool-Calling LoRA

This model is a LoRA adapter fine-tuned from Qwen/Qwen3-0.6B for financial tool calling. It is designed to convert natural-language requests about company fundamentals into structured JSON function calls.

The model does not directly answer financial questions. Instead, it maps a user request into a valid call to a fundamentals retrieval function, including companies, requested metrics, and year ranges.

Intended Use

This model is intended for use inside an agentic financial assistant or tool-routing system. A future orchestrator agent can call this model when a user asks for financial fundamentals data.

Example user request:

Show the revenue and free cash flow for Apple and Microsoft from 2018 to 2023.

Expected model output:

{
  "action": "call",
  "function": "get_fundamentals",
  "arguments": {
    "queries": [
      {
        "symbols": ["Apple", "Microsoft"],
        "metrics": ["Revenue", "Free Cash Flow"],
        "start_year": 2018,
        "end_year": 2023
      }
    ]
  }
}

Training

The adapter was trained with LoRA on a synthetic financial tool-calling dataset. The dataset contains user-style financial requests paired with deterministic JSON completions. The training objective is to produce valid JSON tool calls while preserving:

• Company names
• Requested financial metrics
• Start and end years
• Company-specific metric assignments

Limitations

This model is an initial research/prototype model. It should not be used as a financial advisor and does not retrieve live financial data by itself. The generated JSON should be validated before execution by downstream tools.