A shared language for AI-agent evaluation

EIO-Agents 0.8.0 is the semantic layer between evaluation producers and consumers. The Evaluation Intelligence Ontology (EIO 0.6.0) gives claims, evidence, risks and decisions shared names and rules. The Portable Evaluation Record (PER 2.1.0) carries the result in a versioned, verifiable format.

How EIO and PER fit together

An evaluation framework such as ProofAgent Harness observes an agent and supplies a source bundle. EIO-Agents maps observations to EIO predicates, derives supported findings and scores, and produces a PER. A consumer can validate its structure and, with authorized source access, verify citations and the record digest.

What this means for industry

A common vocabulary and portable record make evaluation results easier to compare, audit and exchange. EIO 0.6.0 includes domain modules and 30 provisional framework, regulatory and taxonomy mapping targets. These mappings identify where evidence may be relevant; they do not establish legal applicability, compliance or certification.

Missing inputs can cause scores to be withheld. The release recommendation remains separate from a score. Verification checks consistency with source evidence; a human reviewer still decides how to use the result.

Reference and documentation

Start with the EIO semantic layer and its core components, use the EIO ontology, framework mappings and versioned schemas reference, then read the PER 2.1.0 guide, sample report and adapters.

EIO-Agents developed by Dr. Fouad Bousetouane · © 2025–2026 ProofAI LLC · Licensed under the Apache License 2.0 · Source on GitHub · support@proofagent.ai