OKF
Screener Space
Article 53 Compliance
Agentic Architecture Experiment

OKF vs. RAG vs.
Agentic CSV

The Open Knowledge Framework (OKF) is a deterministic graph ontology that maps semantic relationships and compliance logic. This live experiment benchmarks OKF against standard RAG and Agentic SQL pipelines to test if graph-traversal yields faster, more accurate EU AI Act verdicts.

Dataset
49 Foundation Models fetched from the Hugging Face Hub.
Screened against the two Article 53 obligations that apply to all GPAI providers, using real model card signals, not just the output license.
Methods
Standard RAG
Dumps raw CSV text into the LLM context.
Agentic CSV
LLM writes SQL to query a flat database.
OKF
LLM traverses semantic relationships in a JSON-LD graph.
02 / AGENT WORKSPACE
Standard RAG ⓘ -
📄
Waiting to start…
Dumps raw text into the LLM context.
Result
Agentic CSV ⓘ -
Waiting to start…
LLM writes SQL to query a flat database.
Result
OKF ⓘ -
M
S
A
Waiting to start…
LLM traverses semantic relationships in a JSON-LD graph.
Result