Sentence-Transformers embedding base
The captured configuration identifies BertModel with sentence-transformers support for embedding workflows.
Open Source Model Profile · dragonkue
multilingual-e5-small-ko-v2
multilingual-e5-small-ko-v2 is a 0.12B-parameter BERT embedding model from dragonkue. According to the model card, it targets Korean retrieval with 384-dimensional vectors.
- Publisher
- dragonkue
- Task
- sentence-similarity
- Model type
- bert
- License
- apache-2.0
- Library
- sentence-transformers
- Publication status
- Approved for indexing
Model overview
multilingual-e5-small-ko-v2 is published by dragonkue as a sentence-similarity model. The captured configuration identifies BertModel, and Safetensors metadata reports about 0.12B parameters. According to the model card, it finetunes intfloat/multilingual-e5-small for Korean retrieval with 384-dimensional embeddings.
Recorded capabilities
Korean retrieval focus
According to the model card, fine-tuning on Korean query-passage pairs targets Korean retrieval performance.
384-dimensional output
According to the model card, sentences and paragraphs map to a 384-dimensional dense vector space.
Model Soup construction
According to the model card, a 6:4 weighted merge combines the Korean-specialized checkpoint with the base multilingual model.
GISTEmbedLoss training
According to the model card, training uses clustered in-batch negatives with GISTEmbedLoss and margin after MNR-only loss reduced performance.
Use cases in the source record
- Korean semantic search, similarity, paraphrase mining, classification, and clustering using 384-dimensional embeddings.
- Retrieval evaluation work consistent with the listed Ko-StrategyQA, AutoRAG, MIRACL, and PublicHealthQA dataset descriptions.
Limitations and unknowns
- The publisher states stronger Korean-benchmark performance than the larger e5-base model, but no extracted numeric table is stated here.
- The card lists overlapping epoch figures in its hyperparameter blocks, so exact epoch count should be confirmed from the training configuration.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
Source and provenance
Source: dragonkue/multilingual-e5-small-ko-v2
Captured: Unknown. Processed: 2026-09-07T19:35:55.713752+00:00.
SentenceTransformer based on intfloat/multilingual-e5-small This is a sentence-transformers model finetuned from intfloat/multilingual-e5-small on datasets that include Korean query-passage pairs for improved performance on Korean retrieval tasks. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more. This model is a lightweight Korean retriever, designed for ease of use and strong performance in practical retrieval tasks. It is ideal for running demos or lightweight applications, offering a…
F001F002F003F004F005F006F007F008F009F010F011F014F015F016F017F021F022F023F024F028F029F030F031F032F033