bitext/Mistral-7B-Customer-Support

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Apr 5, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

bitext/Mistral-7B-Customer-Support is a 7 billion parameter Mistral-7B-Instruct-v0.2 model fine-tuned by Bitext for customer support applications. Optimized to answer questions and assist users with support transactions, it leverages a hybrid synthetic dataset for specialized performance. This model is designed as a verticalized solution to streamline the creation of customized customer service chatbots and virtual assistants.

Loading preview...

Model Overview

bitext/Mistral-7B-Customer-Support is a 7 billion parameter model, fine-tuned from mistralai/Mistral-7B-Instruct-v0.2, specifically for the customer support domain. Developed by Bitext, it is optimized to handle various support transactions and answer customer inquiries.

Key Capabilities

  • Customer Support Specialization: Tailored to understand and generate responses for a wide range of customer service topics, including banking needs.
  • Verticalized Fine-tuning: Designed as a foundational step for creating highly customized chatbots and virtual assistants for specific companies, allowing for easier adaptation with additional fine-tuning on proprietary data.
  • Comprehensive Training Data: Fine-tuned on the bitext/Bitext-customer-support-llm-chatbot-training-dataset, which includes 26,872 question/answer pairs across 27 intents and 10 categories (e.g., cancel_order, check_invoice).
  • Efficient Training: Achieved with a learning rate of 0.0002, 1 epoch, and a maximum sequence length of 1024 tokens.

Intended Use Cases

  • First-step Fine-tuning: Ideal for Bitext's two-step approach to LLM fine-tuning for customer support chatbots, virtual assistants, and copilots.
  • Banking Customer Service: Provides fast and accurate answers for banking-related customer inquiries.

Limitations

  • Domain Specificity: Not intended for general conversational purposes and may not perform well outside the customer service domain.
  • Ethical Considerations: Users should be mindful of potential biases from the training data and ensure responsible use, complementing human expertise.