Core concepts

Metrics & scoring

Understand the metric views shared across AIDX results, then open each evaluation module for its method-specific scoring.

AIDX results move from a summary metric to dimensions, categories, and case-level evidence. The meaning, direction, and calculation of each metric depend on the evaluation module.

Read results at a glance

View	What it tells you	Check
Summary score or rate	The overall signal for the selected evaluation method	Direction, scale, and evaluated denominator
Dimension	Which broad risk area drives the result	Differences hidden by the overall average
Category	Which specific behavior or control needs attention	Concentrated weaknesses
Case	The prompt, target response, outcome, and explanation	Whether the evidence supports the aggregate

Metrics by evaluation module

Open the module guide for its metric definitions, direction, breakdowns, and interpretation rules.

Benchmark testing

BenchDX metrics

Understand benchmark scores, safety dimensions, categories, and case evidence.

Robustness testing

RobustDX metrics

Understand robustness results, attack success, attack methods, and failure evidence.

Coming soon

HalluDX metrics

Preview planned hallucination-risk and factual-consistency measures.

Coming soon

AlignDX metrics

Preview planned policy-alignment measures and violation evidence.