Navigating Uncertainties: How GenAI Developers Document Their Models on Open-Source Platforms

Developers lack clarity on documentation requirements: Despite guidelines from platforms like Hugging Face, GenAI developers interviewed in the study reported significant uncertainty about what specific content should be included in model documentation, suggesting that existing resources may not provide sufficiently clear or actionable guidance.

Reporting technical details remains challenging: The research identified a second layer of uncertainty around how to effectively communicate key model components, indicating that developers struggle not just with what to document, but with translating complex technical information into useful documentation formats.

Responsibility gaps create accountability problems: Developers expressed confusion about who should be responsible for different aspects of documentation—whether that's model creators, platform maintainers, or downstream users—revealing unclear accountability structures that could undermine responsible AI practices.

Community norms and infrastructure need strengthening: The study recommends that platforms, policymakers, and researchers work together to establish clearer community standards, build better evaluation tools for documentation quality, and explicitly define roles and responsibilities to address the documented uncertainties.

Current guidelines are insufficient in practice: The gap between existing responsible AI documentation guidelines and actual developer practices suggests that well-intentioned policies alone are not enough; implementation support and clearer expectations are needed to translate principles into consistent, meaningful documentation.