WizardLMTeam/WizardLM-13B-V1.0
WizardLM-13B-V1.0 is a 13 billion parameter instruction-tuned language model developed by WizardLMTeam. It is part of the WizardLM family, which includes models specialized in coding (WizardCoder) and mathematics (WizardMath). This version is fine-tuned for general conversational AI, achieving a 6.35 MT-Bench score and 75.31% on AlpacaEval, making it suitable for detailed and polite assistant-like responses.
Loading preview...
WizardLM-13B-V1.0 Overview
WizardLM-13B-V1.0 is a 13 billion parameter instruction-tuned language model developed by the WizardLMTeam. It is designed to provide helpful, detailed, and polite answers in conversational settings. This model is part of a broader family that includes specialized variants like WizardCoder for code generation and WizardMath for mathematical reasoning, though WizardLM-13B-V1.0 focuses on general-purpose conversational AI.
Key Capabilities
- General Conversational AI: Optimized for engaging in chat-based interactions, providing comprehensive responses.
- Instruction Following: Fine-tuned to accurately follow user instructions and generate relevant output.
- Assistant-like Behavior: Designed to act as a helpful and polite assistant, suitable for various query types.
Performance Highlights
- Achieved an MT-Bench score of 6.35, indicating its capability in multi-turn conversations.
- Scored 75.31% on AlpacaEval, demonstrating its alignment with human preferences for instruction following.
- Recorded 24.0 pass@1 on HumanEval, showing a foundational ability in code generation, though less specialized than WizardCoder models.
Good For
- Chatbots and Virtual Assistants: Ideal for applications requiring detailed and polite conversational responses.
- General-purpose Language Tasks: Suitable for a wide range of tasks where instruction following and comprehensive answers are crucial.
- Research and Development: A solid base model for further fine-tuning on specific conversational datasets, particularly under a non-commercial license.