Human-aligned AI Model Cards with Weighted Hierarchy Architecture

Pengyue Yang, Haolin Jin, Qingwen Zeng, Jiawen Wen, Harry Rao, Huaming Chen

Abstract:

The proliferation of Large Language Models (LLMs) has led to a burgeoning ecosystem of specialized, domain-specific models. While this rapid growth accelerates innovation, it has simultaneously created significant challenges in effective model discovery and adoption. Users often struggle to navigate this fragmented landscape due to inconsistent, incomplete, and imbalanced documentation across platforms.
Existing documentation frameworks, such as Model Cards and FactSheets, have advanced efforts toward standardize reporting, which are still often static, largely qualitative, and not always well suited for rigorous cross-model comparison lacking quantitative mechanisms. This gap exacerbates model underutilization and impedes responsible adoption. To address these gaps, we introduce the Comprehensive Responsible AI Model Card Framework (CRAI-MCF), a novel framework that transitions from static disclosures to actionable, human-aligned documentation. Grounded in Value Sensitive Design, CRAI-MCF is built upon an empirical analysis of 240 open-source projects, distilling 217 parameters into an eight-module, value-aligned architecture. Our framework introduces a quantitative sufficiency criterion to operationalize documentation sufficiency and support more rigorous cross-model documentation comparison within a unified scheme. By integrating technical, ethical, and operational dimensions, CRAI-MCF provides a structured and extensible template for authoring, organizing, and maintaining LLM documentation, with scoring used as a lightweight aid for prioritization and sufficiency checking.