Krishan-1890/qwen2.5-3b-risk-extractor
Krishan-1890/qwen2.5-3b-risk-extractor is a 3.1 billion parameter model, fine-tuned from Qwen/Qwen2.5-3B-Instruct using QLoRA (4-bit NF4). It specializes in extracting material risk factors from financial disclosure passages, outputting them in a structured JSON format with a 12-category taxonomy, severity, affected metric, and verbatim evidence. This model is designed for financial analysis applications requiring automated risk identification from text, offering a context length of 32768 tokens.
Loading preview...
Overview of Krishan-1890/qwen2.5-3b-risk-extractor
This model is a specialized fine-tune of the Qwen/Qwen2.5-3B-Instruct base model, developed by Krishan-1890. It leverages QLoRA (4-bit NF4 quantization with r=16, alpha=32) for efficient adaptation. The primary function of this 3.1 billion parameter model is to identify and extract material risk factors from financial disclosure texts.
Key Capabilities
- Structured Risk Extraction: Extracts risk factors into a detailed JSON format.
- Categorization: Utilizes a 12-category taxonomy for classifying identified risks.
- Detailed Output: Provides severity, affected metrics, and verbatim evidence quotes for each extracted risk.
- Financial Domain Focus: Specifically trained for analyzing financial disclosure passages.
- Efficient Fine-tuning: Uses QLoRA for parameter-efficient fine-tuning, with adapters merged for direct use.
Good for
- Automated analysis of financial reports and disclosures.
- Identifying and categorizing material risk factors in large text datasets.
- Applications requiring structured data output for risk assessment.
- Developers looking for a compact model (3.1B parameters) with a specialized financial NLP capability.