Reliable model evaluation is essential for practical application of large-scale language models. Existing benchmark-based evaluation methods rely on fixed reference answers, limiting their ability to capture important qualitative aspects of generated responses. To address these shortcomings, this paper proposes SPEED, an integrated evaluation framework that leverages expert feature experts to perform comprehensive and descriptive analysis of model output. SPEED actively integrates expert feedback across multiple dimensions, including hallucination detection, toxicity assessment, and lexical-contextual appropriateness. Experimental results demonstrate that SPEED achieves robust and consistent evaluation performance across diverse domains and datasets. Furthermore, by utilizing relatively small and efficient expert models, SPEED demonstrates superior resource efficiency compared to large-scale evaluation tools. These results demonstrate that SPEED significantly improves fairness and interpretability in LLM evaluation and presents a promising alternative to existing evaluation methodologies.
