Abstention Classifier

The AbstentionClassifier in deepeval is a categorical LLM-as-a-judge that determines whether your LLM application said it did not know when the answer was absent from the provided context, answered from context that supported it, or fabricated one.

A good dataset for this includes some deliberately unanswerable questions, with retrieval_context populated so the judge can see what information the application actually had to work with.

Labels

• abstained: the provided context does not contain the answer and the response says so instead of answering.
• answered: the provided context contains the answer and the response gives it.
• fabricated: the provided context does not contain the answer but the response gives one anyway.

Usage

Populate retrieval_context (or context) so the judge can tell whether the answer was available.

from deepeval.classifiers import AbstentionClassifier
from deepeval.test_case import LLMTestCase
from deepeval import evaluate

classifier = AbstentionClassifier()

test_case = LLMTestCase(
    input="What is the refund window for enterprise plans?",
    actual_output="I don't have that information in our policy documents. Please contact your account manager.",
    retrieval_context=["Standard plans can be refunded within 30 days of purchase."],
    expected_labels={classifier.name: "abstained"},
)

evaluate(test_cases=[test_case], classifiers=[classifier])

There are FIVE optional parameters when creating a AbstentionClassifier:

• [Optional] model: a string specifying which of OpenAI's GPT models to use, OR any custom LLM model of type DeepEvalBaseLLM. Defaulted to gpt-5.4.
• [Optional] include_reason: a boolean which when set to True, includes a reason for the chosen label. Defaulted to True.
• [Optional] allow_none: a boolean which when set to True, lets the classifier return no label when none of them fit (surfaced as label=None, with a reason). When False, the closest label is always chosen. Defaulted to False.
• [Optional] async_mode: a boolean which when set to True, enables concurrent execution within the classify() method. Defaulted to True.
• [Optional] classification_template: a subclass of ClassifierTemplate used to override the default prompts. Defaulted to ClassifierTemplate.

As a standalone

You can also run the AbstentionClassifier on a single test case as a standalone, one-off execution. classify() returns the label and stores the result on the instance:

...

label = classifier.classify(test_case)
print(classifier.label, classifier.reason)

How Is It Calculated?

The AbstentionClassifier is a one-shot LLM-as-a-judge: a single call to your evaluation model with the test case content and the labels above, returning the chosen label and a reason. The result is compared against the test case's expected_labels entry, if any, to decide pass or fail.