Model Overview

giraffe176/Open_Maid_Samantha_Hermes_Orca is a 7 billion parameter language model developed by giraffe176. It was created using the SLERP merge method to combine the capabilities of four distinct base models:

• cognitivecomputations/samantha-1.1-westlake-7b
• NeverSleep/Noromaid-7B-0.4-DPO
• teknium/OpenHermes-2.5-Mistral-7B
• Open-Orca/Mistral-7B-OpenOrca

This merging strategy aims to synthesize the strengths of these diverse models into a single, more capable entity.

Performance Benchmarks

The model's performance has been evaluated on the Open LLM Leaderboard, demonstrating solid results across several key metrics. It achieved an average score of 68.81.

Key benchmark results include:

• AI2 Reasoning Challenge (25-Shot): 66.81
• HellaSwag (10-Shot): 85.83
• MMLU (5-Shot): 64.58
• TruthfulQA (0-shot): 53.91
• Winogrande (5-shot): 80.35
• GSM8k (5-shot): 61.41

Detailed evaluation results are available on the Hugging Face Open LLM Leaderboard and its specific details page.

Merge Configuration

The model was constructed through a multi-stage SLERP merge process, as detailed in the provided YAML configuration. This involved merging pairs of models (Samantha with Noromaid, and OpenHermes with OpenOrca) into intermediate 'workspace' models, which were then merged to produce the final model. The configuration specifies bfloat16 dtype for the merge operations.