Model Card: Phinance-Phi-3.5-mini-instruct-finance-v0.2

Overview

Phinance-Phi-3.5-mini-instruct-finance-v0.2 is a fine-tuned mini language model specifically designed for financial tasks, instruction following, and multi-turn conversations. It leverages the Phinance Dataset to excel in finance-specific reasoning, question answering, and lightweight expert applications. The model is based on the phi-3.5-mini architecture, optimized for instruction-based workflows in the financial domain.

Key Features

• Finance-Focused Reasoning: Handles complex tasks like portfolio analysis, market trends, and financial question answering.
• Instruction Following: Trained for fine-grained instruction-based tasks within the financial sector.
• Multi-Turn Conversations: Designed to handle context-aware dialogue with a focus on finance.
• RAG-Compatible: Supports retrieval-augmented generation (RAG) through the use of data tokens (<|data|>) to integrate external data seamlessly.
• Lightweight Architecture: Efficient for deployment on resource-constrained environments while maintaining robust performance.

Training Data

The model was fine-tuned on the Phinance Dataset, a curated subset of financial content. The dataset includes multi-turn conversations formatted in PHI style, with financial relevance scored using advanced keyword matching.

Dataset Highlights:

• Topics: Market trends, investment strategies, financial analysis, and more.
• Format: Conversations in PHI format, including data tokens (<|data|>) for RAG use cases.
• Filtering: High-quality finance-relevant content scored and selected using advanced methods.

Supported Tasks

1. Financial QA: Answer complex questions about market analysis, financial terms, or investment strategies.
2. Multi-Turn Conversations: Engage in context-aware dialogues about financial topics.
3. Instruction Following: Execute finance-specific instructions and prompts with precision.
4. Lightweight Finance Domain Expert Agent: Serve as an efficient, finance-focused assistant for lightweight systems.
5. Retrieval-Augmented Generation (RAG): Seamlessly integrate external data using the <|data|> token for enhanced responses.

Usage

This model is ideal for:

• Financial advisors or assistants
• Chatbots and conversational agents
• Financial QA systems
• Lightweight domain-specific applications for finance

Help Here

Like my work? Want to see more? Custom request? Message me on discord: joseph.flowers.ra Donate here: https://buymeacoffee.com/josephgflowers

How to Use

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Josephgflowers/Phinance-Phi-3.5-mini-instruct-finance-v0.2"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Example usage
inputs = tokenizer("Explain the difference between stocks and bonds.", return_tensors="pt")
outputs = model.generate(**inputs)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))


Limitations and Considerations

    Niche Knowledge: While proficient in financial topics, the model may not perform as well on general-purpose tasks.
    Bias: Data filtering may introduce biases toward certain financial sectors or topics.
    Hallucinations: As with any language model, responses should be verified for accuracy in critical applications.

Model Details

    Base Model: phi-3.5-mini
    Fine-Tuned Dataset: Phinance Dataset
    Version: v0.2
    Parameters: Mini-sized architecture for efficient performance
    Training Framework: Hugging Face Transformers

License

This model is licensed under the Apache 2.0 license.
Citation

If you use this model, please cite:

@model{phinance_phi_3_5_mini_instruct_v0_2,
  title={Phinance-Phi-3.5-mini-instruct-finance-v0.2},
  author={Joseph G. Flowers},
  year={2025},
  url={https://huggingface.co/Josephgflowers/Phinance-Phi-3.5-mini-instruct-finance-v0.2}
}