Coherence
Coherence checks whether an output is logically organised and internally consistent. It mirrors the RAGAS aspect critic for coherence and returns Jev's probability that the text holds together.
What the evaluator does
Jev returns p(yes) for “logically organised, flows well, no internal contradictions” over the output alone.
`output` is logically organised, flows well, and contains no internal contradictions.
The question, verbatim
This is the question the API sends for the example below, generated from the same code path the playground and API use. Jev sees the request fields as state and returns a probability for each outcome. Nothing is generated, so there is nothing to parse.
`output` is logically organised, flows well, and contains no internal contradictions.
Reach for it when
- Long-form generation: reports, emails, articles.
- Detecting self-contradiction after retrieval merges conflicting sources.
Not the right tool when
- Short factual answers where structure is not the question.
Watch out for
- Fluent but wrong text is coherent; this is a form check, not a truth check.
What to send
score 0–1 (p of the good outcome), label “nn% yes”, confidence |p − 0.5| × 2, probabilities yes/no, passed at 0.5.
{
"evaluators": [
"coherence"
],
"output": "Refunds take 5–7 days. Refunds are instant. Contact us if the instant refund takes longer than a week."
}Low p(coherent): the text contradicts itself.
Aggregate with
One result per item is a fact; a dataset of them is a metric. These are the aggregations that fit this evaluator's output shape.
- Is coherence a correctness check?
- No. It judges organisation and internal consistency. A well-organised wrong answer is coherent.