Model Overview

Gille/StrangeMerges_48-7B-dare_ties is a 7 billion parameter language model developed by Gille. It is a product of merging three different models: Gille/StrangeMerges_46-7B-dare_ties, AurelPx/Percival_01-7b-slerp, and Gille/StrangeMerges_47-7B-dare_ties. This merge was performed using the dare_ties method via LazyMergekit, with Locutusque/Hercules-4.0-Mistral-v0.2-7B serving as its base model.

Key Capabilities & Performance

This model demonstrates a balanced performance across various benchmarks, achieving an average score of 57.89 on the Open LLM Leaderboard. Its specific benchmark results include:

• AI2 Reasoning Challenge (25-Shot): 60.92
• HellaSwag (10-Shot): 80.13
• MMLU (5-Shot): 49.51
• TruthfulQA (0-shot): 65.55
• Winogrande (5-shot): 75.85
• GSM8k (5-shot): 15.39

These scores indicate its proficiency in common sense reasoning, factual recall, and language understanding, while also highlighting areas like mathematical reasoning (GSM8k) where there might be room for improvement.

Usage

The model supports standard text generation tasks and can be easily integrated using the Hugging Face transformers library. It is configured for bfloat16 dtype, optimizing for performance on compatible hardware.