Overview

Finch is a 7 billion parameter language model developed by antiven0m, created through a SLERP merge of two distinct 7B models: macadeliccc/WestLake-7B-v2-laser-truthy-dpo and SanjiWatsuki/Kunoichi-DPO-v2-7B. This merging technique aims to combine the strengths of both base models, resulting in a versatile and capable language model.

Key Capabilities & Performance

Finch demonstrates solid performance across a range of benchmarks, achieving an average score of 73.78 on the Open LLM Leaderboard. Notable scores include:

• AI2 Reasoning Challenge (25-Shot): 71.59
• HellaSwag (10-Shot): 87.87
• MMLU (5-Shot): 64.81
• TruthfulQA (0-shot): 67.96
• Winogrande (5-shot): 84.14
• GSM8k (5-shot): 66.34

Detailed evaluation results are available on the Open LLM Leaderboard.

Recommended Usage

For optimal performance, Finch is designed to be used with the ChatML format. Specific sampler settings are recommended:

• Temperature: 1.2
• Min P: 0.2
• Smoothing Factor: 0.2

Quantized versions, including EXL2 and GGUF formats, are also available for efficient deployment.