ARTFEED — Contemporary Art Intelligence

LLM-FACETS: Open-Source Framework for Auditing AI Transparency

ai-technology · 2026-06-01

LLM-FACETS (LLM FActuality Cross-EvaluaTion System) is an open-source framework designed to evaluate Large Language Models for factual grounding, epistemic calibration, and methodological reproducibility. It features a browser-accessible interface and plugin architecture, targeting three practitioner profiles: technical experts, domain experts, and compliance officers. These profiles align with stakeholder categories in the EU AI Act and NIST AI Risk Management Framework. The framework makes data flows explicit and includes deterministic metrics like BLEU and ROUGE. It addresses barriers faced by non-technical practitioners, such as programming requirements and privacy concerns with cloud-hosted platforms. The paper is available on arXiv under ID 2605.31167.

Key facts

  • LLM-FACETS is an open-source framework for evaluating LLM transparency and accountability.
  • It assesses factual grounding, epistemic calibration, and methodological reproducibility.
  • The framework has a browser-accessible interface and plugin architecture.
  • It targets three practitioner profiles: technical experts, domain experts, compliance officers.
  • Profiles mirror stakeholder categories in the EU AI Act and NIST AI Risk Management Framework.
  • Data flows are made explicit in the architecture.
  • Deterministic metrics include BLEU and ROUGE.
  • The paper is on arXiv with ID 2605.31167.

Entities

Institutions

  • arXiv
  • EU AI Act
  • NIST AI Risk Management Framework

Sources