Misted-7B Model Summary

Misted-7B is a 7 billion parameter language model developed by Walmart-the-bag, built upon the Mistral architecture. This model is a product of a spherical linear interpolation (slerp) merge, combining the teknium/OpenHermes-2-Mistral-7B and Open-Orca/Mistral-7B-SlimOrca models with a merge parameter t of 0.32. This merging strategy aims to synthesize the beneficial characteristics of both base models into a single, cohesive unit.

Key Capabilities & Performance

Misted-7B demonstrates solid performance across various benchmarks, as evaluated on the Open LLM Leaderboard. Its average score of 66.94 indicates proficiency in diverse tasks. Specific benchmark results include:

• AI2 Reasoning Challenge (25-Shot): 63.65
• HellaSwag (10-Shot): 84.14
• MMLU (5-Shot): 63.94
• TruthfulQA (0-shot): 52.00
• Winogrande (5-shot): 78.30
• GSM8k (5-shot): 59.59

These scores highlight its ability in reasoning, common sense, factual recall, and mathematical problem-solving. The model uses an Alpaca-style prompt format for instruction following.

Good For

• General-purpose instruction following tasks.
• Applications requiring a balance of reasoning and common sense.
• Scenarios where a 7B parameter model offers a good trade-off between performance and computational resources.