KielEmbed-Pro

A flagship, high-performance dense embedding model fine-tuned for enterprise retrieval and semantic similarity tasks.

This is a sentence-transformers model fine-tuned from BAAI/bge-large-en-v1.5. It maps sentences and paragraphs into a high-fidelity 1024-dimensional dense vector space optimized for cross-domain retrieval.

Model Details

Model Description

• Model Type: Sentence Transformer / Dense Retrieval Backbone
• Base Model: BAAI/bge-large-en-v1.5
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity
• Supported Modality: Text (Multi-Domain Enterprise & General Corpus)

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face Hub: kiel/KielEmbed-Pro

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}}, 'module_output_name': 'token_embeddings', 'architecture': 'BertModel'})
  (1): Pooling({'embedding_dimension': 1024, 'pooling_mode': 'cls', 'include_prompt': True})
  (2): Normalize({})
)
UsageDirect Usage (Sentence Transformers)First, install the Sentence Transformers library:Bashpip install -U sentence-transformers
Then load the model and run inference:Pythonfrom sentence_transformers import SentenceTransformer

# Load your custom cloud-hosted flagship embedder
model = SentenceTransformer("kiel/KielEmbed-Pro")

# Run inference
sentences = [
    'Many conservatives have staunchly opposed condom programs , saying they send the wrong message and encourage and enable teens to have sex before marriage .',
    'Some conservative groups have staunchly opposed such programs , saying they send the wrong message and in effect encourage and enable teens to have sex before marriage .',
    "It 's just a matter of time , said Frank McDonald , of the University of Maryland .",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
Training DetailsTraining DatasetKiel Multi-Domain CorpusSize: 20,000 training samplesColumns: anchor and positiveApproximate Token Statistics (First 1,000 samples):Anchor: Min: 10 tokens | Mean: 27.34 tokens | Max: 51 tokensPositive: Min: 10 tokens | Mean: 27.39 tokens | Max: 51 tokensLoss Function: MultipleNegativesRankingLoss with parameters:JSON{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "gather_across_devices": false,
    "directions": [
        "query_to_doc"
    ],
    "partition_mode": "joint",
    "hardness_mode": null,
    "hardness_strength": 0.0
}
Training HyperparametersPer Device Train Batch Size: 2Gradient Accumulation Steps: 16 (Effective batch size = 32)Learning Rate: 2e-05Number of Epochs: 1Warmup Steps: 0.1Mixed Precision: FP16 EnabledGradient Checkpointing: EnabledOptimizer: adamw_torch_fusedTraining LogsEpochStepTraining Loss0.16500.01770.321000.01350.481500.00710.642000.00740.82500.00940.963000.0097Total Training Time: 1.2 hoursFramework VersionsPython: 3.12.13Sentence Transformers: 5.4.1Transformers: 5.0.0PyTorch: 2.10.0+cu128Accelerate: 1.13.0Datasets: 5.0.0Tokenizers: 0.22.2CitationBibTeXSentence TransformersCode snippet@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "[https://arxiv.org/abs/1908.10084](https://arxiv.org/abs/1908.10084)",
}
MultipleNegativesRankingLossCode snippet@misc{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={[https://arxiv.org/abs/1807.03748](https://arxiv.org/abs/1807.03748)},
}