bunnycore/Tulu-3.1-8B-SuperNova
bunnycore/Tulu-3.1-8B-SuperNova is an 8 billion parameter language model based on the Llama 3.1 architecture, created by merging three distinct Llama 3.1-based models using a linear merge method. This model integrates capabilities from MedIT-SUN, Tulu-3, and SuperNova-Lite, aiming for a balanced performance across various tasks. It is designed for general-purpose applications, leveraging its merged components to offer broad utility.
Loading preview...
Overview
bunnycore/Tulu-3.1-8B-SuperNova is an 8 billion parameter language model built upon the Llama 3.1 architecture. It was developed using a linear merge method, combining three specialized Llama 3.1-based models: meditsolutions/Llama-3.1-MedIT-SUN-8B, allenai/Llama-3.1-Tulu-3-8B, and arcee-ai/Llama-3.1-SuperNova-Lite. This merging strategy aims to consolidate the strengths of its constituent models into a single, versatile offering.
Key Capabilities & Performance
The model's performance has been evaluated on the Open LLM Leaderboard, with detailed results available here. Key metrics include:
- Avg. Score: 30.94
- IFEval (0-Shot): 81.94
- BBH (3-Shot): 32.50
- MATH Lvl 5 (4-Shot): 24.32
- MMLU-PRO (5-shot): 31.27
These scores indicate its proficiency in instruction following, general reasoning, and some mathematical tasks, reflecting the combined training of its merged components.
When to Use This Model
This model is suitable for developers seeking a Llama 3.1-based model that integrates diverse capabilities from multiple fine-tunes. Its balanced performance across various benchmarks suggests it can be a good candidate for:
- General-purpose conversational AI.
- Instruction-following tasks.
- Applications requiring a blend of reasoning and factual recall.