Vicuna-13B-V1.1 Overview

Vicuna-13B-V1.1 is an open-source chatbot model developed by the Vicuna team, a collaboration of researchers from UC Berkeley, CMU, Stanford, and UC San Diego. It is built upon the LLaMA-13B architecture and has been fine-tuned using user-shared conversations collected from ShareGPT. This version, released in April 2023, focuses on enhancing compatibility and model quality.

Key Capabilities

• Chatbot Functionality: Designed for conversational AI tasks, leveraging its fine-tuning on diverse user interactions.
• LLaMA-based Architecture: Benefits from the robust transformer architecture of the LLaMA model family.
• Improved Tokenization: Features a refactored tokenization scheme where the End-Of-Sentence (EOS) token "</s>" is used as the separator, simplifying generation stop criteria and improving compatibility with various libraries.
• Enhanced Fine-tuning: Incorporates fixes in the supervised fine-tuning loss computation, contributing to better overall model quality.

Good For

• Research on Large Language Models: Ideal for academic and independent researchers exploring advancements in LLMs and chatbot development.
• Hobbyist Projects: Suitable for enthusiasts in natural language processing, machine learning, and artificial intelligence looking to experiment with and deploy conversational AI.
• Chatbot Prototyping: Can be used to build and test conversational agents, benefiting from its instruction-following capabilities derived from ShareGPT data.

For more detailed information, refer to the official Vicuna project page: https://vicuna.lmsys.org/