HWERI/Llama2-7b-openorca-mc-v2: Multiple Choice Optimized Llama2 Variant

HWERI/Llama2-7b-openorca-mc-v2 is a 7 billion parameter language model built upon the Llama2 architecture. Its key differentiator lies in its specialized fine-tuning process, which involved a 10,000-sample subset of the OpenOrca dataset, specifically curated for multiple-choice questions, combined with 6,000 samples from the ShareGPT4 dataset.

Key Capabilities & Performance

This model demonstrates strong performance in multiple-choice question answering scenarios, as reflected in its Open LLM Leaderboard evaluation results:

• ARC (25-shot): Achieved 55.55
• HellaSwag (10-shot): Scored 81.26
• MMLU (5-shot): Reached 48.3
• Winogrande (5-shot): Posted 72.85

While excelling in multiple-choice and common sense reasoning, its performance on mathematical reasoning (GSM8K) and reading comprehension (DROP) is lower, indicating a focused optimization for specific task types.

Ideal Use Cases

This model is particularly well-suited for applications requiring robust multiple-choice question answering, such as:

• Educational assessment tools
• Quiz generation and solving
• Fact-checking systems where answers are presented as options
• Any scenario where the primary task involves selecting the correct answer from a predefined set of choices.