Datasets:
Languages:
English
Size:
1K - 10K
ArXiv:
Tags:
enoki
hallucination-detection
factuality
open-information-extraction
natural-language-inference
long-form-question-answering
License:
Place paper link below GitHub repository
Browse files
README.md
CHANGED
|
@@ -36,8 +36,8 @@ natural generator and Wikipedia-popularity distribution. It is intended for
|
|
| 36 |
research on long-form question answering, factuality, hallucination detection,
|
| 37 |
and evidence-based verification.
|
| 38 |
|
| 39 |
-
- Paper: [Enoki: Efficient Multi-Level Hallucination Detection](https://arxiv.org/abs/2609.00581)
|
| 40 |
- Code: [s-nlp/Enoki](https://github.com/s-nlp/Enoki)
|
|
|
|
| 41 |
- Original release: [Harvard Dataverse](https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/6TN4ZM)
|
| 42 |
|
| 43 |
## Loading the dataset
|
|
|
|
| 36 |
research on long-form question answering, factuality, hallucination detection,
|
| 37 |
and evidence-based verification.
|
| 38 |
|
|
|
|
| 39 |
- Code: [s-nlp/Enoki](https://github.com/s-nlp/Enoki)
|
| 40 |
+
- Paper: [Enoki: Efficient Multi-Level Hallucination Detection](https://arxiv.org/abs/2609.00581)
|
| 41 |
- Original release: [Harvard Dataverse](https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/6TN4ZM)
|
| 42 |
|
| 43 |
## Loading the dataset
|