Model Overview
CNCL-Penn-State/CrPO-sft-llama-3.1-8b-instruct is an 8 billion parameter language model built upon the Llama-3.1 architecture. It has been specifically supervised-finetuned (SFT) using the MuCE-SFT dataset. This fine-tuning process is derived from the research presented in the "Creative Preference Optimization" paper, indicating a specialized focus on creative generation and understanding.
Key Capabilities
- Creative Preference Optimization: The model's training on the MuCE-SFT dataset suggests a strong capability in tasks related to creative content generation and evaluating creative preferences.
- Instruction Following: As an instruction-tuned model, it is designed to accurately interpret and execute user prompts.
- Extended Context Window: Supports a context length of 32768 tokens, allowing for processing and generating longer, more complex creative inputs and outputs.
Use Cases
This model is particularly well-suited for applications requiring:
- Creative Text Generation: Generating diverse and contextually relevant creative content.
- Creative Content Evaluation: Potentially assisting in tasks that involve assessing or optimizing creative outputs based on learned preferences.
- Research in Creativity: Serving as a base model for further research into computational creativity and preference modeling, as outlined in the Creative Preference Optimization paper.