flowaicom/Flow-Judge-v0.1

Flow-Judge-v0.1 by flowaicom is a compact 3.8 billion parameter model based on the Phi-3.5-mini instruct architecture, designed for customizable LLM system evaluations. It excels at providing high-quality, structured evaluations with qualitative feedback and numerical scores across various scales (Pass/Fail, 3-Likert, 5-Likert). This model is optimized for cost-effective and rapid evaluation of generative AI products using custom rubrics, achieving performance comparable to larger models in specific evaluation benchmarks.