mteb 2.0.5__py3-none-any.whl → 2.1.1__py3-none-any.whl

mteb/tasks/classification/mya/myanmar_news.py

@@ -9,7 +9,7 @@ class MyanmarNews(AbsTaskClassification):
             "path": "mteb/MyanmarNews",
             "revision": "644419f24bc820bbf8af24e0b4714a069812e0a3",
         },
-        description="The Myanmar News dataset on Hugging Face contains news articles in Burmese. It is designed for tasks such as text classification, sentiment analysis, and language modeling. The dataset includes a variety of news topics in 4 categorie, providing a rich resource for natural language processing applications involving Burmese which is a low resource language.",
+        description="The Myanmar News dataset on Hugging Face contains news articles in Burmese. It is designed for tasks such as text classification, sentiment analysis, and language modeling. The dataset includes a variety of news topics in 4 categories, providing a rich resource for natural language processing applications involving Burmese which is a low resource language.",
         reference="https://huggingface.co/datasets/myanmar_news",
         type="Classification",
         category="t2c",
@@ -45,7 +45,7 @@ class MyanmarNewsV2(AbsTaskClassification):
             "path": "mteb/myanmar_news",
             "revision": "475b43ffbdb5138ad67a01a2c860bc7db502f3c5",
         },
-        description="""The Myanmar News dataset on Hugging Face contains news articles in Burmese. It is designed for tasks such as text classification, sentiment analysis, and language modeling. The dataset includes a variety of news topics in 4 categorie, providing a rich resource for natural language processing applications involving Burmese which is a low resource language.
+        description="""The Myanmar News dataset on Hugging Face contains news articles in Burmese. It is designed for tasks such as text classification, sentiment analysis, and language modeling. The dataset includes a variety of news topics in 4 categories, providing a rich resource for natural language processing applications involving Burmese which is a low resource language.
         This version corrects errors found in the original data. For details, see [pull request](https://github.com/embeddings-benchmark/mteb/pull/2900)""",
         reference="https://huggingface.co/datasets/myanmar_news",
         type="Classification",

mteb/tasks/classification/nld/__init__.py

@@ -2,8 +2,24 @@ from .dutch_book_review_sentiment_classification import (
     DutchBookReviewSentimentClassification,
     DutchBookReviewSentimentClassificationV2,
 )
+from .dutch_cola_classification import DutchColaClassification
+from .dutch_government_bias_classification import DutchGovernmentBiasClassification
+from .dutch_news_articles_classification import DutchNewsArticlesClassification
+from .dutch_sarcastic_headlines_classification import (
+    DutchSarcasticHeadlinesClassification,
+)
+from .iconclass_classification import IconclassClassification
+from .open_tender_classification import OpenTenderClassification
+from .vaccin_chat_nl_classification import VaccinChatNLClassification
 
 __all__ = [
     "DutchBookReviewSentimentClassification",
     "DutchBookReviewSentimentClassificationV2",
+    "DutchColaClassification",
+    "DutchGovernmentBiasClassification",
+    "DutchNewsArticlesClassification",
+    "DutchSarcasticHeadlinesClassification",
+    "IconclassClassification",
+    "OpenTenderClassification",
+    "VaccinChatNLClassification",
 ]

mteb/tasks/classification/nld/dutch_cola_classification.py

@@ -0,0 +1,38 @@
+from mteb.abstasks.classification import AbsTaskClassification
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class DutchColaClassification(AbsTaskClassification):
+    samples_per_label = 128
+    metadata = TaskMetadata(
+        name="DutchColaClassification",
+        description="Dutch CoLA is a corpus of linguistic acceptability for Dutch.",
+        reference="https://huggingface.co/datasets/GroNLP/dutch-cola",
+        dataset={
+            "path": "clips/mteb-nl-dutch-cola",
+            "revision": "2269ed7d95d8abaab829f1592b4b2047372e9f81",
+        },
+        type="Classification",
+        category="t2c",
+        modalities=["text"],
+        date=("2024-03-01", "2024-05-01"),
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="f1",
+        domains=["Written"],
+        task_subtypes=["Linguistic acceptability"],
+        license="not specified",  # specified as unknown
+        annotations_creators="expert-annotated",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation=r"""
+@misc{gronlp_2024,
+  author = {Bylinina, Lisa and Abdi, Silvana and Brouwer, Hylke and Elzinga, Martine and Gunput, Shenza and Huisman, Sem and Krooneman, Collin and Poot, David and Top, Jelmer and Weideman, Cain},
+  doi = { 10.57967/hf/3825 },
+  publisher = { Hugging Face },
+  title = { {Dutch-CoLA (Revision 5a4196c)} },
+  url = { https://huggingface.co/datasets/GroNLP/dutch-cola },
+  year = {2024},
+}
+""",
+    )

mteb/tasks/classification/nld/dutch_government_bias_classification.py

@@ -0,0 +1,37 @@
+from mteb.abstasks.classification import AbsTaskClassification
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class DutchGovernmentBiasClassification(AbsTaskClassification):
+    samples_per_label = 32
+    metadata = TaskMetadata(
+        name="DutchGovernmentBiasClassification",
+        description="The Dutch Government Data for Bias Detection (DGDB) is a dataset sourced from the Dutch House of Representatives and annotated for bias by experts",
+        reference="https://dl.acm.org/doi/pdf/10.1145/3696410.3714526",
+        dataset={
+            "path": "clips/mteb-nl-dutch-government-bias-detection",
+            "revision": "bf5e20ee2d3ce2e24e4de50f5dd8573e0e0e2fec",
+        },
+        type="Classification",
+        category="t2c",
+        modalities=["text"],
+        date=("2019-10-04", "2019-10-04"),
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="f1",
+        domains=["Written", "Government"],
+        task_subtypes=[],
+        license="cc-by-nc-sa-4.0",
+        annotations_creators="expert-annotated",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation=r"""
+@inproceedings{de2025detecting,
+  author = {de Swart, Milena and Den Hengst, Floris and Chen, Jieying},
+  booktitle = {Proceedings of the ACM on Web Conference 2025},
+  pages = {5034--5044},
+  title = {Detecting Linguistic Bias in Government Documents Using Large language Models},
+  year = {2025},
+}
+""",
+    )

mteb/tasks/classification/nld/dutch_news_articles_classification.py

@@ -0,0 +1,30 @@
+from mteb.abstasks.classification import AbsTaskClassification
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class DutchNewsArticlesClassification(AbsTaskClassification):
+    metadata = TaskMetadata(
+        name="DutchNewsArticlesClassification",
+        dataset={
+            "path": "clips/mteb-nl-news-articles-cls",
+            "revision": "0a7227d31f85c5676be92767f8df5405ea93de54",
+        },
+        description="This dataset contains all the articles published by the NOS as of the 1st of January 2010. The "
+        "data is obtained by scraping the NOS website. The NOS is one of the biggest (online) news "
+        "organizations in the Netherlands.",
+        reference="https://www.kaggle.com/datasets/maxscheijen/dutch-news-articles",
+        type="Classification",
+        category="t2c",
+        modalities=["text"],
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="f1",
+        date=("2009-11-01", "2010-01-01"),
+        domains=["Written", "News"],
+        task_subtypes=["Topic classification"],
+        license="cc-by-nc-sa-4.0",
+        annotations_creators="derived",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation="",
+    )

mteb/tasks/classification/nld/dutch_sarcastic_headlines_classification.py

@@ -0,0 +1,36 @@
+from mteb.abstasks.classification import AbsTaskClassification
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class DutchSarcasticHeadlinesClassification(AbsTaskClassification):
+    metadata = TaskMetadata(
+        name="DutchSarcasticHeadlinesClassification",
+        description="This dataset contains news headlines of two Dutch news websites. All sarcastic headlines were "
+        "collected from the Speld.nl (the Dutch equivalent of The Onion) whereas all 'normal' headlines "
+        "were collected from the news website Nu.nl.",
+        reference="https://www.kaggle.com/datasets/harrotuin/dutch-news-headlines",
+        dataset={
+            "path": "clips/mteb-nl-sarcastic-headlines",
+            "revision": "7e520e36394795859583f84f81fcb97de915d05a",
+        },
+        type="Classification",
+        category="t2c",
+        modalities=["text"],
+        date=("2019-01-01", "2020-01-01"),
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="f1",
+        domains=["News", "Written", "Fiction"],
+        task_subtypes=[],
+        license="cc0-1.0",
+        annotations_creators="derived",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation="""""",
+    )
+
+    def dataset_transform(self):
+        for split in self.dataset:
+            self.dataset[split] = self.dataset[split].rename_columns(
+                {"headline": "text", "is_sarcastic": "label"}
+            )

mteb/tasks/classification/nld/iconclass_classification.py

@@ -0,0 +1,41 @@
+from mteb.abstasks.classification import AbsTaskClassification
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class IconclassClassification(AbsTaskClassification):
+    samples_per_label = 32
+    metadata = TaskMetadata(
+        name="IconclassClassification",
+        description="Iconclass is an iconographic thesaurus, which is widely used in the digital heritage domain to "
+        "describe subjects depicted in artworks. The task is to classify the first layer of Iconclass",
+        reference="https://dl.acm.org/doi/pdf/10.1145/3575865",
+        dataset={
+            "path": "clips/mteb-nl-iconclass-cls",
+            "revision": "1cd02f1579dab39fedc95de8cc15fd620557a9f2",
+        },
+        type="Classification",
+        category="t2c",
+        modalities=["text"],
+        date=("2020-01-01", "2020-05-01"),
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="f1",
+        domains=["Written", "Fiction"],
+        task_subtypes=[],
+        license="cc-by-nc-sa-4.0",
+        annotations_creators="expert-annotated",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation=r"""
+@article{banar2023transfer,
+  author = {Banar, Nikolay and Daelemans, Walter and Kestemont, Mike},
+  journal = {ACM Journal on Computing and Cultural Heritage},
+  number = {2},
+  pages = {1--16},
+  publisher = {ACM New York, NY},
+  title = {Transfer learning for the visual arts: The multi-modal retrieval of iconclass codes},
+  volume = {16},
+  year = {2023},
+}
+""",
+    )

mteb/tasks/classification/nld/open_tender_classification.py

@@ -0,0 +1,38 @@
+from mteb.abstasks.classification import AbsTaskClassification
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class OpenTenderClassification(AbsTaskClassification):
+    metadata = TaskMetadata(
+        name="OpenTenderClassification",
+        dataset={
+            "path": "clips/mteb-nl-opentender-cls-pr",
+            "revision": "9af5657575a669dc18c7f897a67287ff7d1a0c65",
+        },
+        description="This dataset contains Belgian and Dutch tender calls from OpenTender in Dutch",
+        reference="https://arxiv.org/abs/2509.12340",
+        type="Classification",
+        category="t2c",
+        modalities=["text"],
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="f1",
+        date=("2025-08-01", "2025-08-10"),
+        domains=["Government", "Written"],
+        task_subtypes=[],
+        license="cc-by-4.0",
+        annotations_creators="human-annotated",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation=r"""
+@misc{banar2025mtebnle5nlembeddingbenchmark,
+  archiveprefix = {arXiv},
+  author = {Nikolay Banar and Ehsan Lotfi and Jens Van Nooten and Cristina Arhiliuc and Marija Kliocaite and Walter Daelemans},
+  eprint = {2509.12340},
+  primaryclass = {cs.CL},
+  title = {MTEB-NL and E5-NL: Embedding Benchmark and Models for Dutch},
+  url = {https://arxiv.org/abs/2509.12340},
+  year = {2025},
+}
+""",
+    )

mteb/tasks/classification/nld/vaccin_chat_nl_classification.py

@@ -0,0 +1,46 @@
+from mteb.abstasks.classification import AbsTaskClassification
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class VaccinChatNLClassification(AbsTaskClassification):
+    metadata = TaskMetadata(
+        name="VaccinChatNLClassification",
+        description="VaccinChatNL is a Flemish Dutch FAQ dataset on the topic of COVID-19 vaccinations in Flanders.",
+        reference="https://huggingface.co/datasets/clips/VaccinChatNL",
+        dataset={
+            "path": "clips/VaccinChatNL",
+            "revision": "bd27d0058bea2ad52470d9072a3b5da6b97c1ac3",
+        },
+        type="Classification",
+        category="t2c",
+        modalities=["text"],
+        date=("2022-01-01", "2022-09-01"),
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="f1",
+        domains=["Spoken", "Web"],
+        task_subtypes=[],
+        license="cc-by-4.0",
+        annotations_creators="expert-annotated",
+        dialect=[],
+        sample_creation="created",
+        bibtex_citation=r"""
+@inproceedings{buhmann-etal-2022-domain,
+  address = {Gyeongju, Republic of Korea},
+  author = {Buhmann, Jeska and De Bruyn, Maxime and Lotfi, Ehsan and Daelemans, Walter},
+  booktitle = {Proceedings of the 29th International Conference on Computational Linguistics},
+  month = oct,
+  pages = {3539--3549},
+  publisher = {International Committee on Computational Linguistics},
+  title = {Domain- and Task-Adaptation for {V}accin{C}hat{NL}, a {D}utch {COVID}-19 {FAQ} Answering Corpus and Classification Model},
+  url = {https://aclanthology.org/2022.coling-1.312},
+  year = {2022},
+}
+""",
+    )
+
+    def dataset_transform(self):
+        for split in self.dataset:
+            self.dataset[split] = self.dataset[split].rename_columns(
+                {"sentence1": "text"}
+            )

mteb/tasks/classification/tha/wongnai_reviews_classification.py

@@ -5,7 +5,7 @@ from mteb.abstasks.task_metadata import TaskMetadata
 class WongnaiReviewsClassification(AbsTaskClassification):
     metadata = TaskMetadata(
         name="WongnaiReviewsClassification",
-        description="Wongnai features over 200,000 restaurants, beauty salons, and spas across Thailand on its platform, with detailed information about each merchant and user reviews. In this dataset there are 5 classes corressponding each star rating",
+        description="Wongnai features over 200,000 restaurants, beauty salons, and spas across Thailand on its platform, with detailed information about each merchant and user reviews. In this dataset there are 5 classes corresponding each star rating",
         reference="https://github.com/wongnai/wongnai-corpus",
         dataset={
             "path": "Wongnai/wongnai_reviews",

mteb/tasks/classification/ukr/ukr_formality_classification.py

@@ -10,7 +10,7 @@ class UkrFormalityClassification(AbsTaskClassification):
         trainslating English GYAFC data.
         English data source: https://aclanthology.org/N18-1012/
         Translation into Ukrainian language using model: https://huggingface.co/facebook/nllb-200-distilled-600M
-        Additionally, the dataset was balanced, witha labels: 0 - informal, 1 - formal.
+        Additionally, the dataset was balanced, with labels: 0 - informal, 1 - formal.
         """,
         dataset={
             "path": "ukr-detect/ukr-formality-dataset-translated-gyafc",
@@ -61,7 +61,7 @@ class UkrFormalityClassificationV2(AbsTaskClassification):
         trainslating English GYAFC data.
         English data source: https://aclanthology.org/N18-1012/
         Translation into Ukrainian language using model: https://huggingface.co/facebook/nllb-200-distilled-600M
-        Additionally, the dataset was balanced, witha labels: 0 - informal, 1 - formal.
+        Additionally, the dataset was balanced, with labels: 0 - informal, 1 - formal.
 
         This version corrects errors found in the original data. For details, see [pull request](https://github.com/embeddings-benchmark/mteb/pull/2900)""",
         dataset={

mteb/tasks/clustering/__init__.py

@@ -5,6 +5,7 @@ from .fra import *
 from .jpn import *
 from .kor import *
 from .multilingual import *
+from .nld import *
 from .nob import *
 from .pol import *
 from .rom import *

mteb/tasks/clustering/nld/__init__.py

@@ -0,0 +1,17 @@
+from .dutch_news_articles_clustering_p2p import DutchNewsArticlesClusteringP2P
+from .dutch_news_articles_clustering_s2s import DutchNewsArticlesClusteringS2S
+from .iconclass_clustering_s2s import IconclassClusteringS2S
+from .open_tender_clustering_p2p import OpenTenderClusteringP2P
+from .open_tender_clustering_s2s import OpenTenderClusteringS2S
+from .vabb_clustering_p2p import VABBClusteringP2P
+from .vabb_clustering_s2s import VABBClusteringS2S
+
+__all__ = [
+    "DutchNewsArticlesClusteringP2P",
+    "DutchNewsArticlesClusteringS2S",
+    "IconclassClusteringS2S",
+    "OpenTenderClusteringP2P",
+    "OpenTenderClusteringS2S",
+    "VABBClusteringP2P",
+    "VABBClusteringS2S",
+]

mteb/tasks/clustering/nld/dutch_news_articles_clustering_p2p.py

@@ -0,0 +1,37 @@
+from mteb.abstasks.clustering import AbsTaskClustering
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class DutchNewsArticlesClusteringP2P(AbsTaskClustering):
+    max_fraction_of_documents_to_embed = 1.0
+    metadata = TaskMetadata(
+        name="DutchNewsArticlesClusteringP2P",
+        dataset={
+            "path": "clips/mteb-nl-news-articles-cls",
+            "revision": "0a7227d31f85c5676be92767f8df5405ea93de54",
+        },
+        description="This dataset contains all the articles published by the NOS as of the 1st of January 2010. The "
+        "data is obtained by scraping the NOS website. The NOS is one of the biggest (online) news "
+        "organizations in the Netherlands.",
+        reference="https://www.kaggle.com/datasets/maxscheijen/dutch-news-articles",
+        type="Clustering",
+        category="t2c",
+        modalities=["text"],
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="v_measure",
+        date=("2009-11-01", "2010-01-01"),
+        domains=["Written", "News"],
+        task_subtypes=[],
+        license="cc-by-nc-sa-4.0",
+        annotations_creators="derived",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation="",
+    )
+
+    def dataset_transform(self):
+        for split in self.dataset:
+            self.dataset[split] = self.dataset[split].rename_columns(
+                {"label": "labels", "text": "sentences"}
+            )

mteb/tasks/clustering/nld/dutch_news_articles_clustering_s2s.py

@@ -0,0 +1,37 @@
+from mteb.abstasks.clustering import AbsTaskClustering
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class DutchNewsArticlesClusteringS2S(AbsTaskClustering):
+    max_fraction_of_documents_to_embed = 1.0
+    metadata = TaskMetadata(
+        name="DutchNewsArticlesClusteringS2S",
+        dataset={
+            "path": "clips/mteb-nl-news-articles-cls",
+            "revision": "0a7227d31f85c5676be92767f8df5405ea93de54",
+        },
+        description="This dataset contains all the articles published by the NOS as of the 1st of January 2010. The "
+        "data is obtained by scraping the NOS website. The NOS is one of the biggest (online) news "
+        "organizations in the Netherlands.",
+        reference="https://www.kaggle.com/datasets/maxscheijen/dutch-news-articles",
+        type="Clustering",
+        category="t2c",
+        modalities=["text"],
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="v_measure",
+        date=("2009-11-01", "2010-01-01"),
+        domains=["Written", "News"],
+        task_subtypes=[],
+        license="cc-by-nc-sa-4.0",
+        annotations_creators="derived",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation="",
+    )
+
+    def dataset_transform(self):
+        for split in self.dataset:
+            self.dataset[split] = self.dataset[split].rename_columns(
+                {"label": "labels", "title": "sentences"}
+            )

mteb/tasks/clustering/nld/iconclass_clustering_s2s.py

@@ -0,0 +1,47 @@
+from mteb.abstasks.clustering import AbsTaskClustering
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class IconclassClusteringS2S(AbsTaskClustering):
+    max_fraction_of_documents_to_embed = 1.0
+    metadata = TaskMetadata(
+        name="IconclassClusteringS2S",
+        dataset={
+            "path": "clips/mteb-nl-iconclass-cls",
+            "revision": "1cd02f1579dab39fedc95de8cc15fd620557a9f2",
+        },
+        description="Iconclass is an iconographic thesaurus, which is widely used in the digital heritage domain to "
+        "describe subjects depicted in artworks. The task is to classify the first layer of Iconclass",
+        reference="https://dl.acm.org/doi/pdf/10.1145/3575865",
+        type="Clustering",
+        category="t2c",
+        modalities=["text"],
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="v_measure",
+        date=("2009-11-01", "2010-01-01"),
+        domains=["Written", "Fiction"],
+        task_subtypes=[],
+        license="cc-by-nc-sa-4.0",
+        annotations_creators="derived",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation=r"""
+@article{banar2023transfer,
+  author = {Banar, Nikolay and Daelemans, Walter and Kestemont, Mike},
+  journal = {ACM Journal on Computing and Cultural Heritage},
+  number = {2},
+  pages = {1--16},
+  publisher = {ACM New York, NY},
+  title = {Transfer learning for the visual arts: The multi-modal retrieval of iconclass codes},
+  volume = {16},
+  year = {2023},
+}
+""",
+    )
+
+    def dataset_transform(self):
+        for split in self.dataset:
+            self.dataset[split] = self.dataset[split].map(
+                lambda ex: {"labels": ex["label"], "sentences": ex["text"]}
+            )

mteb/tasks/clustering/nld/open_tender_clustering_p2p.py

@@ -0,0 +1,51 @@
+from mteb.abstasks.clustering import AbsTaskClustering
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class OpenTenderClusteringP2P(AbsTaskClustering):
+    max_fraction_of_documents_to_embed = 1.0
+    metadata = TaskMetadata(
+        name="OpenTenderClusteringP2P",
+        dataset={
+            "path": "clips/mteb-nl-opentender-cls-pr",
+            "revision": "9af5657575a669dc18c7f897a67287ff7d1a0c65",
+        },
+        description="This dataset contains all the articles published by the NOS as of the 1st of January 2010. The "
+        "data is obtained by scraping the NOS website. The NOS is one of the biggest (online) news "
+        "organizations in the Netherlands.",
+        reference="https://arxiv.org/abs/2509.12340",
+        type="Clustering",
+        category="t2c",
+        modalities=["text"],
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="v_measure",
+        date=("2025-08-01", "2025-08-10"),
+        domains=["Government", "Written"],
+        task_subtypes=[],
+        license="cc-by-nc-sa-4.0",
+        annotations_creators="derived",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation=r"""
+@misc{banar2025mtebnle5nlembeddingbenchmark,
+  archiveprefix = {arXiv},
+  author = {Nikolay Banar and Ehsan Lotfi and Jens Van Nooten and Cristina Arhiliuc and Marija Kliocaite and Walter Daelemans},
+  eprint = {2509.12340},
+  primaryclass = {cs.CL},
+  title = {MTEB-NL and E5-NL: Embedding Benchmark and Models for Dutch},
+  url = {https://arxiv.org/abs/2509.12340},
+  year = {2025},
+}
+""",
+    )
+
+    def dataset_transform(self):
+        # reuse the dataset for classification
+        for split in self.dataset:
+            self.dataset[split] = self.dataset[split].map(
+                lambda ex: {
+                    "labels": ex["label"],
+                    "sentences": ex["text"],
+                }
+            )

mteb/tasks/clustering/nld/open_tender_clustering_s2s.py

@@ -0,0 +1,41 @@
+from mteb.abstasks.clustering import AbsTaskClustering
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class OpenTenderClusteringS2S(AbsTaskClustering):
+    max_fraction_of_documents_to_embed = 1.0
+    metadata = TaskMetadata(
+        name="OpenTenderClusteringS2S",
+        dataset={
+            "path": "clips/mteb-nl-opentender-clst-s2s-pr",
+            "revision": "ad86cf1813d130e17dda0092d1c4f2c664e68d0c",
+        },
+        description="This dataset contains all the articles published by the NOS as of the 1st of January 2010. The "
+        "data is obtained by scraping the NOS website. The NOS is one of the biggest (online) news "
+        "organizations in the Netherlands.",
+        reference="https://arxiv.org/abs/2509.12340",
+        type="Clustering",
+        category="t2c",
+        modalities=["text"],
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="v_measure",
+        date=("2025-08-01", "2025-08-10"),
+        domains=["Government", "Written"],
+        task_subtypes=[],
+        license="cc-by-nc-sa-4.0",
+        annotations_creators="derived",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation=r"""
+@misc{banar2025mtebnle5nlembeddingbenchmark,
+  archiveprefix = {arXiv},
+  author = {Nikolay Banar and Ehsan Lotfi and Jens Van Nooten and Cristina Arhiliuc and Marija Kliocaite and Walter Daelemans},
+  eprint = {2509.12340},
+  primaryclass = {cs.CL},
+  title = {MTEB-NL and E5-NL: Embedding Benchmark and Models for Dutch},
+  url = {https://arxiv.org/abs/2509.12340},
+  year = {2025},
+}
+""",
+    )

mteb/tasks/clustering/nld/vabb_clustering_p2p.py

@@ -0,0 +1,51 @@
+from mteb.abstasks.clustering import AbsTaskClustering
+from mteb.abstasks.task_metadata import TaskMetadata
+
+
+class VABBClusteringP2P(AbsTaskClustering):
+    max_fraction_of_documents_to_embed = 1.0
+    metadata = TaskMetadata(
+        name="VABBClusteringP2P",
+        dataset={
+            "path": "clips/mteb-nl-vabb-cls",
+            "revision": "544acc2e46909eab2b49962b043a18b9c9772770",
+        },
+        description="This dataset contains the fourteenth edition of the Flemish Academic Bibliography for the Social "
+        "Sciences and Humanities (VABB-SHW), a database of academic publications from the social sciences "
+        "and humanities authored by researchers affiliated to Flemish universities (more information). "
+        "Publications in the database are used as one of the parameters of the Flemish performance-based "
+        "research funding system",
+        reference="https://zenodo.org/records/14214806",
+        type="Clustering",
+        category="t2c",
+        modalities=["text"],
+        eval_splits=["test"],
+        eval_langs=["nld-Latn"],
+        main_score="v_measure",
+        date=("2009-11-01", "2010-01-01"),
+        domains=["Academic", "Written"],
+        task_subtypes=[],
+        license="cc-by-nc-sa-4.0",
+        annotations_creators="derived",
+        dialect=[],
+        sample_creation="found",
+        bibtex_citation=r"""
+@dataset{aspeslagh2024vabb,
+  author = {Aspeslagh, Pieter and Guns, Raf and Engels, Tim C. E.},
+  doi = {10.5281/zenodo.14214806},
+  publisher = {Zenodo},
+  title = {VABB-SHW: Dataset of Flemish Academic Bibliography for the Social Sciences and Humanities (edition 14)},
+  url = {https://doi.org/10.5281/zenodo.14214806},
+  year = {2024},
+}
+""",
+    )
+
+    def dataset_transform(self):
+        for split in self.dataset:
+            self.dataset[split] = self.dataset[split].map(
+                lambda ex: {
+                    "labels": ex["org_discipline"],
+                    "sentences": f"{ex['title']}\n{ex['abstract']}",
+                }
+            )