🏗️ Building on HF
Tom Aarsen
AI & ML interests
NLP: text embeddings, information retrieval, named entity recognition, few-shot text classification
Recent Activity
liked a model about 1 hour ago
papr-ai/papr-embed-code-1-0.6b liked a model about 3 hours ago
model-collapse/pixmoji liked a Space about 17 hours ago
HuggingFaceSwarms/pybot-msgboardOrganizations
Encoder comparison after equivalent retrieval fine-tuning
ModernBERT, Ettin and UltraBERT: 54 retrieval models, frozen data and results. Validation-selected models first. Matched training, one seed.
-
tomaarsen/encoder-comparison-20261014
38.2 MB -
tomaarsen/encoder-comparison-20261014
Updated • 311 -
tomaarsen/modernbert-base-dense-msmarco32k-1e-4-20261014
Sentence Similarity • 0.1B • Updated • 16 -
tomaarsen/modernbert-base-multivector-msmarco32k-3e-4-20261014
Feature Extraction • 0.1B • Updated • 15
Matryoshka Embedding Models
https://huggingface.co/blog/matryoshka
-
BEE-spoke-data/bert-plus-L8-v1.0-syntheticSTS-4k
Sentence Similarity • 88.1M • Updated • 55 • 5 -
aspire/acge_text_embedding
Sentence Similarity • 0.3B • Updated • 451 • 150 -
dunzhang/stella-mrl-large-zh-v3.5-1792d
Sentence Similarity • 0.3B • Updated • 534 • 51 -
NeuML/pubmedbert-base-embeddings-matryoshka
Sentence Similarity • 0.1B • Updated • 2.84k • • 23
State-of-the-Art NER models - General purpose
-
tomaarsen/span-marker-bert-base-fewnerd-fine-super
Token Classification • 0.1B • Updated • 3.45k • 15 -
tomaarsen/span-marker-roberta-large-fewnerd-fine-super
Token Classification • 0.4B • Updated • 105 • 14 -
tomaarsen/span-marker-mbert-base-multinerd
Token Classification • 0.2B • Updated • 465 • 72 -
tomaarsen/span-marker-roberta-large-ontonotes5
Token Classification • 0.4B • Updated • 85 • 15
State-of-the-Art NER models - Acronyms
State-of-the-Art NER models - Tagalog
SpanMarker NER Models
SpanMarker NER models for various domains
SetFit models
Reranker Models for GooAQ
https://huggingface.co/blog/train-reranker
-
tomaarsen/reranker-ModernBERT-large-gooaq-bce
Text Ranking • 0.4B • Updated • 1.06k • • 9 -
tomaarsen/reranker-NeoBERT-gooaq-bce
Text Ranking • 0.2B • Updated • 38 • 2 -
tomaarsen/reranker-ModernBERT-base-gooaq-bce
Text Ranking • 0.1B • Updated • 89 • 3 -
tomaarsen/reranker-MiniLM-L12-gooaq-bce
Text Ranking • 33.4M • Updated • 34 • 1
Training with Prompts
See the Training with Prompts documentation for more details: https://sbert.net/examples/training/prompts/README.html
Reranker Models for MS MARCO
State-of-the-Art NER models - Biomedical domain
State-of-the-Art NER models - Keyphrases
State-of-the-Art NER models - Organizations
-
nbroad/span-marker-roberta-large-orgs-v1
Token Classification • 0.4B • Updated • 22 • 2 -
tomaarsen/span-marker-bert-base-orgs
Token Classification • Updated • 40 • 1 -
nbroad/span-marker-xdistil-l12-h384-orgs-v3
Token Classification • 33.4M • Updated • 12 -
tomaarsen/span-marker-bert-small-orgs
Token Classification • Updated • 15
Strong & Small Rerankers
Trained on MS MARCO using the a similar strategy as https://huggingface.co/cross-encoder/ms-marco-MiniLM-L12-v2, except with the Ettin base models
SetFitABSA models
-
tomaarsen/setfit-absa-bge-small-en-v1.5-restaurants-aspect
Text Classification • Updated • 515 • 4 -
tomaarsen/setfit-absa-bge-small-en-v1.5-restaurants-polarity
Text Classification • Updated • 497 -
tomaarsen/setfit-absa-paraphrase-mpnet-base-v2-restaurants-aspect
Text Classification • 0.1B • Updated • 95 • 1 -
tomaarsen/setfit-absa-paraphrase-mpnet-base-v2-restaurants-polarity
Text Classification • 0.1B • Updated • 86
Qwen3 Rerankers converted to Sequence Classification
Reranker Models for GooAQ
https://huggingface.co/blog/train-reranker
-
tomaarsen/reranker-ModernBERT-large-gooaq-bce
Text Ranking • 0.4B • Updated • 1.06k • • 9 -
tomaarsen/reranker-NeoBERT-gooaq-bce
Text Ranking • 0.2B • Updated • 38 • 2 -
tomaarsen/reranker-ModernBERT-base-gooaq-bce
Text Ranking • 0.1B • Updated • 89 • 3 -
tomaarsen/reranker-MiniLM-L12-gooaq-bce
Text Ranking • 33.4M • Updated • 34 • 1
Encoder comparison after equivalent retrieval fine-tuning
ModernBERT, Ettin and UltraBERT: 54 retrieval models, frozen data and results. Validation-selected models first. Matched training, one seed.
-
tomaarsen/encoder-comparison-20261014
38.2 MB -
tomaarsen/encoder-comparison-20261014
Updated • 311 -
tomaarsen/modernbert-base-dense-msmarco32k-1e-4-20261014
Sentence Similarity • 0.1B • Updated • 16 -
tomaarsen/modernbert-base-multivector-msmarco32k-3e-4-20261014
Feature Extraction • 0.1B • Updated • 15
Training with Prompts
See the Training with Prompts documentation for more details: https://sbert.net/examples/training/prompts/README.html
Matryoshka Embedding Models
https://huggingface.co/blog/matryoshka
-
BEE-spoke-data/bert-plus-L8-v1.0-syntheticSTS-4k
Sentence Similarity • 88.1M • Updated • 55 • 5 -
aspire/acge_text_embedding
Sentence Similarity • 0.3B • Updated • 451 • 150 -
dunzhang/stella-mrl-large-zh-v3.5-1792d
Sentence Similarity • 0.3B • Updated • 534 • 51 -
NeuML/pubmedbert-base-embeddings-matryoshka
Sentence Similarity • 0.1B • Updated • 2.84k • • 23
Reranker Models for MS MARCO
State-of-the-Art NER models - General purpose
-
tomaarsen/span-marker-bert-base-fewnerd-fine-super
Token Classification • 0.1B • Updated • 3.45k • 15 -
tomaarsen/span-marker-roberta-large-fewnerd-fine-super
Token Classification • 0.4B • Updated • 105 • 14 -
tomaarsen/span-marker-mbert-base-multinerd
Token Classification • 0.2B • Updated • 465 • 72 -
tomaarsen/span-marker-roberta-large-ontonotes5
Token Classification • 0.4B • Updated • 85 • 15
State-of-the-Art NER models - Biomedical domain
State-of-the-Art NER models - Acronyms
State-of-the-Art NER models - Keyphrases
State-of-the-Art NER models - Tagalog
State-of-the-Art NER models - Organizations
-
nbroad/span-marker-roberta-large-orgs-v1
Token Classification • 0.4B • Updated • 22 • 2 -
tomaarsen/span-marker-bert-base-orgs
Token Classification • Updated • 40 • 1 -
nbroad/span-marker-xdistil-l12-h384-orgs-v3
Token Classification • 33.4M • Updated • 12 -
tomaarsen/span-marker-bert-small-orgs
Token Classification • Updated • 15
SpanMarker NER Models
SpanMarker NER models for various domains
Strong & Small Rerankers
Trained on MS MARCO using the a similar strategy as https://huggingface.co/cross-encoder/ms-marco-MiniLM-L12-v2, except with the Ettin base models
SetFit models
SetFitABSA models
-
tomaarsen/setfit-absa-bge-small-en-v1.5-restaurants-aspect
Text Classification • Updated • 515 • 4 -
tomaarsen/setfit-absa-bge-small-en-v1.5-restaurants-polarity
Text Classification • Updated • 497 -
tomaarsen/setfit-absa-paraphrase-mpnet-base-v2-restaurants-aspect
Text Classification • 0.1B • Updated • 95 • 1 -
tomaarsen/setfit-absa-paraphrase-mpnet-base-v2-restaurants-polarity
Text Classification • 0.1B • Updated • 86