clips/e5-large-trm
0103
1---2license: mit3language: nl4base_model:5- intfloat/multilingual-e5-large6pipeline_tag: sentence-similarity7library_name: sentence-transformers8tags:9- transformers10---11# E5-large-trm12This model is a trimmed version of [intfloat/multilingual-e5-large](https://huggingface.co/intfloat/multilingual-e5-large) by [`vocabtrimmer`](https://github.com/asahi417/lm-vocab-trimmer), a tool for trimming vocabulary of language models to compress the model size.13Following table shows a summary of the trimming process.14 15| | intfloat/multilingual-e5-large | clips/e5-large-trm |16|:---------------------------|:-------------------------------|:-------------------|17| parameter_size_full | 559,890,432 | 355,090,432 |18| parameter_size_embedding | 256,002,048 | 51,202,048 |19| vocab_size | 250,002 | 50,002 |20| compression_rate_full | 100.0 | 63.42 |21| compression_rate_embedding | 100.0 | 20.0 |22 23 24Following table shows the parameter used to trim vocabulary.25 26 | language | dataset | dataset_column | dataset_name | dataset_split | target_vocab_size | min_frequency |27|:-----------|:-----------|:-----------------|:---------------|:----------------|--------------------:|----------------:|28| nl | allenai/c4 | text | nl | validation | 50000 | 2 |29 30## Usage31 32Below is an example to encode queries and passages from the MS-MARCO passage ranking dataset.33 34```python35import torch.nn.functional as F36 37from torch import Tensor38from transformers import AutoTokenizer, AutoModel39 40 41def average_pool(last_hidden_states: Tensor,42 attention_mask: Tensor) -> Tensor:43 last_hidden = last_hidden_states.masked_fill(~attention_mask[..., None].bool(), 0.0)44 return last_hidden.sum(dim=1) / attention_mask.sum(dim=1)[..., None]45 46 47# Each input text should start with "query: " or "passage: ".48# For tasks other than retrieval, you can simply use the "query: " prefix.49input_texts = [50 'query: hoeveel eiwitten moet een vrouw eten',51 'query: top definieer',52 "passage: Als algemene richtlijn geldt dat de gemiddelde eiwitbehoefte voor vrouwen van 19 tot 70 jaar volgens de CDC 46 gram per dag bedraagt. Maar, zoals je in deze tabel kunt zien, moet je dit verhogen als je zwanger bent of traint voor een marathon. Bekijk de onderstaande tabel om te zien hoeveel eiwitten je dagelijks zou moeten eten.",53 "passage: Definitie van top voor leerlingen Engels. : 1 het hoogste punt van een berg : de top van een berg. : 2 het hoogste niveau. : 3 een bijeenkomst of reeks bijeenkomsten tussen de leiders van twee of meer regeringen."54]55 56tokenizer = AutoTokenizer.from_pretrained('clips/e5-large-trm')57model = AutoModel.from_pretrained('clips/e5-large-trm')58 59# Tokenize the input texts60batch_dict = tokenizer(input_texts, max_length=512, padding=True, truncation=True, return_tensors='pt')61 62outputs = model(**batch_dict)63embeddings = average_pool(outputs.last_hidden_state, batch_dict['attention_mask'])64 65# normalize embeddings66embeddings = F.normalize(embeddings, p=2, dim=1)67scores = (embeddings[:2] @ embeddings[2:].T) * 10068print(scores.tolist())69```70 71Below is an example for usage with sentence_transformers.72```python73from sentence_transformers import SentenceTransformer74 75# Load the model from Hugging Face76model = SentenceTransformer("clips/e5-large-trm")77 78# Perform inference using encode_query/encode_document for retrieval,79# or encode_query for general purpose embeddings. Prompt prefixes80# are automatically added with these two methods.81queries = [82 'hoeveel eiwitten moet een vrouw eten',83 'top definieer',84]85documents = [86 'Als algemene richtlijn geldt dat de gemiddelde eiwitbehoefte voor vrouwen van 19 tot 70 jaar volgens de CDC 46 gram per dag bedraagt. Maar, zoals je in deze tabel kunt zien, moet je dit verhogen als je zwanger bent of traint voor een marathon. Bekijk de onderstaande tabel om te zien hoeveel eiwitten je dagelijks zou moeten eten.',87 'Definitie van top voor leerlingen Engels. : 1 het hoogste punt van een berg : de top van een berg. : 2 het hoogste niveau. : 3 een bijeenkomst of reeks bijeenkomsten tussen de leiders van twee of meer regeringen.',88]89query_embeddings = model.encode_query(queries)90document_embeddings = model.encode_document(documents)91print(query_embeddings.shape, document_embeddings.shape)92# (2, 1024) (2, 1024)93 94similarities = model.similarity(query_embeddings, document_embeddings)95# tensor([[0.9112, 0.7261],96# [0.7412, 0.9056]])97```98## Benchmark Evaluation99Results on MTEB-NL (models introduced in [our paper](https://arxiv.org/abs/2509.12340) and the best model per size category are highlighted in bold):100 101| Model | Prm | Cls | MLCls | PCls | Rrnk | Rtr | Clust | STS | AvgD | AvgT |102|---------------------------------------|------|----------|----------|----------|----------|----------|----------|----------|----------|----------|103| **Num. Datasets (→)** | | 12 | 3 | 2 | 1 | 12 | 8 | 2 | 40 | |104| **Supervised (small, <100M)** | | | | | | | | | | |105| **e5-small-v2-t2t** | 33M | 53.7 | 38.5 | 74.5 | 85.9 | 45.0 | 24.1 | 74.3 | 46.9 | 56.6 |106| **e5-small-v2-t2t-nl** | 33M | 55.3 | 40.9 | 74.9 | 86.0 | 49.9 | 28.0 | 74.1 | 49.8 | 58.4 |107| **e5-small-trm** | 41M | 56.3 | 43.5 | **76.5** | **87.3** | 53.1 | 28.2 | 74.2 | 51.4 | 59.9 |108| **e5-small-trm-nl** | 41M | **58.2** | **44.7** | 76.0 | 87.1 | **56.0** | **32.2** | **74.6** | **53.8** | **61.3** |109| **Supervised (base, <305M)** | | | | | | | | | | |110| granite-embedding-107m-multilingual | 107M | 53.9 | 41.8 | 70.1 | 84.7 | 50.2 | 29.8 | 68.4 | 49.4 | 57.0 |111| **e5-base-v2-t2t** | 109M | 54.4 | 40.3 | 73.3 | 85.6 | 46.2 | 25.5 | 73.2 | 47.8 | 56.9 |112| **e5-base-v2-t2t-nl** | 109M | 53.9 | 41.5 | 72.5 | 84.0 | 46.4 | 26.9 | 69.3 | 47.8 | 56.3 |113| multilingual-e5-small | 118M | 56.3 | 43.5 | 76.5 | 87.1 | 53.1 | 28.2 | 74.2 | 51.4 | 59.8 |114| paraphrase-multilingual-MiniLM-L12-v2 | 118M | 55.0 | 38.1 | 78.2 | 80.6 | 37.7 | 29.6 | 76.3 | 46.3 | 56.5 |115| **RobBERT-2023-base-ft** | 124M | 58.1 | 44.6 | 72.7 | 84.7 | 51.6 | 32.9 | 68.5 | 52.0 | 59.0 |116| **e5-base-trm** | 124M | 58.1 | 44.4 | 76.7 | 88.3 | 55.8 | 28.1 | 74.9 | 52.9 | 60.9 |117| **e5-base-trm-nl** | 124M | **59.6** | **45.9** | 78.4 | 87.5 | 56.5 | **34.3** | 75.8 | **55.0** | **62.6** |118| potion-multilingual-128M | 128M | 51.8 | 40.0 | 60.4 | 80.3 | 35.7 | 26.1 | 62.0 | 42.6 | 50.9 |119| multilingual-e5-base | 278M | 58.2 | 44.4 | 76.7 | **88.4** | 55.8 | 27.7 | 74.9 | 52.8 | 60.9 |120| granite-embedding-278m-multilingual | 278M | 54.6 | 41.8 | 71.0 | 85.6 | 52.4 | 30.3 | 68.9 | 50.5 | 58.0 |121| paraphrase-multilingual-mpnet-base-v2 | 278M | 58.1 | 40.5 | **81.9** | 82.3 | 41.4 | 30.8 | 79.3 | 49.2 | 59.2 |122| Arctic-embed-m-v2.0 | 305M | 54.4 | 42.6 | 66.6 | 86.2 | 51.8 | 26.5 | 64.9 | 49.1 | 56.1 |123| gte-multilingual-base | 305M | 59.1 | 37.7 | 77.8 | 82.3 | **56.8** | 31.3 | **78.6** | 53.8 | 60.5 |124| **Supervised (large, >305M)** | | | | | | | | | | |125| **e5-large-v2-t2t** | 335M | 55.7 | 41.4 | 75.7 | 86.6 | 49.9 | 25.5 | 74.0 | 49.5 | 58.4 |126| **e5-large-v2-t2t-nl** | 335M | 57.3 | 42.4 | 76.9 | 86.9 | 50.8 | 27.7 | 74.1 | 51.7 | 59.4 |127| **RobBERT-2023-large-ft** | 355M | 59.3 | 45.2 | 68.7 | 82.3 | 48.3 | 31.6 | 70.6 | 51.0 | 58.0 |128| **e5-large-trm** | 355M | 60.2 | 45.4 | 80.3 | 90.3 | 59.0 | 28.7 | 78.8 | 55.1 | 63.3 |129| **e5-large-trm-nl** | 355M | **62.2** | **48.0** | **81.4** | 87.2 | 58.2 | 35.6 | 78.2 | **57.0** | **64.4** |130| multilingual-e5-large | 560M | 60.2 | 45.4 | 80.3 | **90.3** | 59.1 | 29.5 | 78.8 | 55.3 | 63.4 |131| Arctic-embed-l-v2.0 | 568M | 59.3 | 45.2 | 74.2 | 88.2 | 59.0 | 29.8 | 71.7 | 54.3 | 61.1 |132| bge-m3 | 568M | 60.7 | 44.2 | 78.3 | 88.7 | **60.0** | 29.2 | 78.1 | 55.4 | 63.1 |133| jina-embeddings-v3 | 572M | 61.7 | 38.9 | 76.8 | 78.5 | 59.1 | **38.9** | **84.8** | **57.0** | 62.7 |134 135 136 137### Citation Information138 139If you find our paper, benchmark or models helpful, please consider cite as follows:140```latex141@misc{banar2025mtebnle5nlembeddingbenchmark,142 title={MTEB-NL and E5-NL: Embedding Benchmark and Models for Dutch}, 143 author={Nikolay Banar and Ehsan Lotfi and Jens Van Nooten and Cristina Arhiliuc and Marija Kliocaite and Walter Daelemans},144 year={2025},145 eprint={2509.12340},146 archivePrefix={arXiv},147 primaryClass={cs.CL},148 url={https://arxiv.org/abs/2509.12340}, 149}150```151[//]: # (https://arxiv.org/abs/2509.12340)152