The COMPEL Glossary Graph visualizes relationships between framework terminology, showing how concepts interconnect across domains, stages, and pillars. Term nodes cluster by pillar affiliation while cross-references reveal semantic dependencies — for example, how risk appetite connects to control effectiveness, model governance, and assurance requirements. This network representation helps practitioners navigate the framework vocabulary and understand that COMPEL terminology forms a coherent conceptual system rather than isolated definitions.
COMPEL Glossary / ai-experiment
AI experiment
A structured comparison producing evidence for a decision — about a model version, a prompt, a feature set, a retrieval strategy, or a deployment change.
What this means in practice
Spans four modes: offline (static data), online (live traffic), shadow (production traffic but outputs not exposed), and adversarial (adversarial probing). Distinct from general software experimentation because AI experiments must account for statistical variance, data leakage, and distribution shift.
Synonyms
ML experiment , model experiment
See also
- Offline evaluation — Assessment of an AI system against static datasets — training hold-out, validation set, benchmark corpus — without exposure to live user traffic.
- Online evaluation — Assessment of an AI system under live traffic using randomized or sequential experimental designs — A/B test, multi-armed bandit, canary, or interleaving.
- Red-team experiment — An adversarial experiment designed to probe failure modes rather than validate desired behavior — structured, hypothesis-driven exploration of safety bypass, goal mis-specification, jailbreak, and harm.