Hugging Face Overview
Model Overview
Dampfinchen/Llama-3.1-8B-Ultra-Instruct is an 8 billion parameter language model built upon the NousResearch/Meta-Llama-3.1-8B base. It was created using the DARE TIES merge method, combining the strengths of four distinct fine-tuned Llama 3.1 models:
- nbeerbower/llama3.1-gutenberg-8B
- akjindal53244/Llama-3.1-Storm-8B
- nbeerbower/llama3.1-airoboros3.2-QDT-8B
- Sao10K/Llama-3.1-8B-Stheno-v3.4
This merging approach aims to consolidate diverse capabilities into a single, efficient model. The model is designed to be used with the Llama 3 Instruct prompt template.
Performance Highlights
Evaluations on the Open LLM Leaderboard indicate competitive performance for an 8B model. Key metrics include:
- Avg. Score: 28.98
- IFEval (0-Shot): 80.81
- BBH (3-Shot): 32.49
- MMLU-PRO (5-shot): 31.40
Recommended Use Cases
- General Instruction Following: Excels in responding to diverse prompts and instructions.
- Conversational AI: Suitable for chatbots and interactive applications due to its instruction-tuned nature.
- Text Generation: Capable of generating coherent and contextually relevant text across various topics.