Harpy-7B-Model_Stock Overview

Harpy-7B-Model_Stock is a 7 billion parameter language model developed by DreadPoor. It was constructed using a model_stock merge method, combining three distinct base models: Endevor/InfinityRP-v1-7B, macadeliccc/WestLake-7B-v2-laser-truthy-dpo, and abideen/AlphaMonarch-laser. This merging strategy aims to leverage the strengths of its constituent models to achieve robust performance across various tasks.

Key Capabilities & Performance

Evaluated on the Open LLM Leaderboard, Harpy-7B-Model_Stock achieved an average score of 75.51%. Its performance highlights include:

• HellaSwag (10-Shot): 88.72%
• Winogrande (5-shot): 85.24%
• AI2 Reasoning Challenge (25-Shot): 73.21%
• TruthfulQA (0-shot): 71.35%
• GSM8k (5-shot): 69.45%
• MMLU (5-Shot): 65.07%

These scores indicate strong general reasoning, common sense, and question-answering abilities.

When to Use Harpy-7B-Model_Stock

This model is well-suited for applications requiring a capable 7B parameter model with balanced performance across a variety of benchmarks. Its strong results in HellaSwag and Winogrande suggest particular aptitude for tasks involving common sense reasoning and disambiguation. Developers can integrate it into projects for general text generation, summarization, question answering, and other natural language processing tasks where a merged model's diverse strengths are beneficial.