Model Overview
Kukedlc/Neural4gsm8k is a 7 billion parameter language model developed by Kukedlc. It is a merged model, created using the slerp (spherical linear interpolation) method from two base models: Kukedlc/Neural-4-ARC-7B-slerp and Kukedlc/Neural-4-GSM8K-7B-slerp. This merging strategy aims to combine the strengths of its constituent models, particularly in areas related to reasoning and mathematical problem-solving.
Key Capabilities & Performance
This model demonstrates notable performance across several benchmarks, indicating its proficiency in various reasoning tasks:
- Average Score: Achieves an average score of 75.31 on the Open LLM Leaderboard evaluations.
- AI2 Reasoning Challenge (25-Shot): Scores 72.27, highlighting its ability in complex reasoning.
- GSM8k (5-Shot): Attains 73.39, indicating strong mathematical problem-solving skills.
- MMLU (5-Shot): Reaches 64.76, showcasing general knowledge and understanding.
- HellaSwag (10-Shot): Scores 88.45, demonstrating common sense reasoning.
When to Use This Model
Neural4gsm8k is particularly well-suited for applications requiring:
- Mathematical Reasoning: Its strong performance on GSM8k suggests suitability for tasks involving arithmetic and word problems.
- General Reasoning: The high score on the AI2 Reasoning Challenge makes it a good candidate for logical inference and problem-solving.
- Academic and Research Tasks: Its balanced performance across various benchmarks indicates utility in educational or research contexts where diverse reasoning abilities are needed.