KangalKhan-Ruby-7B-Fixed Overview

KangalKhan-Ruby-7B-Fixed is a 7 billion parameter language model developed by Yuma42. It is a merge of two prominent Mistral-7B based models: argilla/CapybaraHermes-2.5-Mistral-7B and argilla/distilabeled-OpenHermes-2.5-Mistral-7B. The merge was performed using the slerp method via LazyMergekit, combining the strengths of both base models to enhance overall performance.

Key Capabilities & Performance

This model is instruction-tuned and designed for conversational AI, as indicated by its ChatML usage example. It demonstrates competitive performance on standard benchmarks, achieving an average score of 68.68 on the Open LLM Leaderboard. Specific benchmark results include:

• AI2 Reasoning Challenge (25-Shot): 67.24
• HellaSwag (10-Shot): 85.22
• MMLU (5-Shot): 63.21
• TruthfulQA (0-shot): 56.49
• Winogrande (5-shot): 77.98
• GSM8k (5-shot): 61.94

Usage and Availability

The model supports a context length of 4096 tokens and is available in bfloat16 precision. GGUF variants are also provided for efficient local deployment, with specific recommendations for Q4_K_S quantization. Developers can easily integrate KangalKhan-Ruby-7B-Fixed into their applications using the Hugging Face transformers library, with a clear example provided for generating responses based on user prompts.