Correctness
Define measurable criteria and representative test cases for this evaluation dimension.
LLM EVALUATION · PARIMI
LLM Evaluation: correctness, relevance, safety and consistency
Evaluate language-model behaviour with explicit criteria, representative datasets and repeatable regression checks rather than relying on fluency alone.
Relevance
Define measurable criteria and representative test cases for this evaluation dimension.
Completeness
Define measurable criteria and representative test cases for this evaluation dimension.
Groundedness
Define measurable criteria and representative test cases for this evaluation dimension.
Consistency
Define measurable criteria and representative test cases for this evaluation dimension.
Safety
Define measurable criteria and representative test cases for this evaluation dimension.
Instruction following
Define measurable criteria and representative test cases for this evaluation dimension.
Regression
Define measurable criteria and representative test cases for this evaluation dimension.
LLM evaluation should be repeatable
PARIMI connects evaluation criteria to datasets, automated execution and release evidence.