kaitchup/Mayonnaise-4in1-01

Overview

Mayonnaise-4in1-01 is a 7 billion parameter causal language model developed by The Kaitchup, built upon the mistralai/Mistral-7B-v0.1 architecture. This model is notable for its creation via the TIES-merging method using mergekit, combining three distinct models: mncai/mistral-7b-dpo-v5, flemmingmiguel/MBX-7B, and BarryFutureman/NeuralTurdusVariant1-7B. This approach aims to leverage the strengths of multiple models into a single, more capable entity.

Key Capabilities & Performance

The model demonstrates strong performance across a range of benchmarks, achieving an average score of 75.19 on the Open LLM Leaderboard. Specific benchmark results include:

• AI2 Reasoning Challenge (25-Shot): 73.46
• HellaSwag (10-Shot): 88.47
• MMLU (5-Shot): 64.95
• TruthfulQA (0-shot): 69.18
• Winogrande (5-shot): 84.14
• GSM8k (5-shot): 70.96

These scores indicate proficiency in reasoning, common sense, language understanding, and mathematical problem-solving. The model's development process, detailed in "The Mayonnaise: Rank First on the Open LLM Leaderboard with TIES-Merging," highlights an effective strategy for creating high-performing models through merging techniques.

When to Use This Model

Mayonnaise-4in1-01 is suitable for applications requiring a general-purpose language model with a strong foundation in reasoning and language tasks. Its balanced performance across various benchmarks makes it a versatile choice for tasks such as:

• General text generation and completion
• Question answering
• Reasoning-intensive tasks
• Applications where a 7B parameter model offers a good balance between performance and computational efficiency.