EnokiQA

EnokiQA is an annotated dataset for fine-grained hallucination detection in long-form question answering. Each example contains a factual question, a no-context LLM answer, verifying Wikipedia context, sentence-grouped factual triples, and per-triple NLI and hallucination probabilities.

Available splits: dev and test.

• Code: s-nlp/Enoki
• Paper: Enoki: Efficient Multi-Level Hallucination Detection

Loading the dataset

pip install datasets

from datasets import load_dataset

dataset = load_dataset("s-nlp/EnokiQA")
dev = dataset["dev"]
test = dataset["test"]

Inspect hallucinated triples using the canonical hall_prob > 0.5 threshold:

example = test[0]

for sentence in example["sentences"]:
    hallucinated = [
        triple
        for triple in sentence["triples"]
        if triple["hall_prob"] > 0.5
    ]
    if hallucinated:
        print(sentence["sentence"])
        for triple in hallucinated:
            print(triple["triplet"], triple["hall_prob"])

Stream a split without downloading it in full:

test = load_dataset("s-nlp/EnokiQA", split="test", streaming=True)

Dataset structure

Field	Type	Description
id	string	Source example identifier
title	string	Wikipedia article title
question	string	Long-form factual question
answer	string	No-context model-generated answer
context	string	Wikipedia evidence used to verify the answer
n_sentences	int32	Number of annotated answer sentences
n_triplets	int32	Total extracted triples across the answer
sentences	list[Sentence]	Sentence-grouped triple annotations

Sentence

Field	Type	Description
sentence_index	int32	Zero-based position in the answer
sentence	string	Sentence text
triples	list[Triple]	Extracted and verified factual triples

Triple

Field	Type	Description
triplet	list[string] of length 3	[subject, predicate, object]
span	list[int32]	Object's [start, end] character span in answer; empty if unmatched
hypothesis	string	Verbalized triple supplied to the NLI model
entailment	float64	Probability that the context entails the hypothesis
neutral	float64	Probability that the context is neutral
contradiction	float64	Probability that the context contradicts the hypothesis
hall_prob	float64	Hallucination probability: neutral + contradiction

A triple is classified as hallucinated when hall_prob > 0.5.

Dataset statistics

Statistic	Dev	Test
Examples	1,995	1,995
Sentences	58,454	57,987
Triples	499,923	492,268
Average triples per answer	250.6	246.8
Median triples per answer	231	226
Mean per-document hallucination rate	77.8%	76.0%

Annotation method

The annotations were produced automatically by a common pipeline for both splits:

1. Answers were split into sentences.
2. GPT-OSS-120B extracted atomic (subject, predicate, object) triples using incremental decomposition.
3. Triple object spans were localized in the generated answer when possible.
4. Each verbalized triple was checked against the Wikipedia context with an NLI model, producing entailment, neutral, and contradiction probabilities.
5. hall_prob was calculated as neutral + contradiction.

Intended use and limitations

• The dataset supports evaluation and threshold tuning for fine-grained, claim-level hallucination detection.
• Triple extraction and NLI labels are automatically generated and can contain extraction, span-localization, or verification errors.
• Wikipedia evidence can be incomplete or temporally outdated; lack of entailment does not always imply a fact is false.
• This compact annotated release does not include generator-model or Wikipedia popularity metadata.

License

Wikipedia-derived content and the annotation layer are distributed under CC BY-SA 4.0. Preserve attribution and share-alike requirements when redistributing derived datasets.

Citation

@misc{rykov2026enokiefficientmultilevelhallucination,
  title         = {Enoki: Efficient Multi-Level Hallucination Detection},
  author        = {Elisei Rykov and Timur Ionov and Nikolay Ivanov and Maksim Savkin and Maksim Makarenko and Alexander Panchenko and Vasily Konovalov and Julia Belikova},
  year          = {2026},
  eprint        = {2609.00581},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CL},
  url           = {https://arxiv.org/abs/2609.00581},
}