pj-mathematician/JobBGE-small-en-v1.5

Job - Job matching BAAI/bge-small-en-v1.5

Top performing model on TalentCLEF 2025 Task A. Use it for multilingual job title matching

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: BAAI/bge-small-en-v1.5
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 384 dimensions
• Similarity Function: Cosine Similarity
• Training Datasets:
  • full_en
  • full_de
  • full_es
  • full_zh
  • mix

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    'Volksvertreter',
    'Parlamentarier',
    'Oberbürgermeister',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Information Retrieval

• Datasets: full_en, full_es, full_de, full_zh, mix_es, mix_de and mix_zh
• Evaluated with InformationRetrievalEvaluator

Metric	full_en	full_es	full_de	full_zh	mix_es	mix_de	mix_zh
cosine_accuracy@1	0.6571	0.1243	0.2956	0.3495	0.4113	0.2943	0.0971
cosine_accuracy@20	0.9905	1.0	0.9212	0.7379	0.7613	0.65	0.3586
cosine_accuracy@50	0.9905	1.0	0.9655	0.8252	0.8523	0.7608	0.4901
cosine_accuracy@100	0.9905	1.0	0.9754	0.8544	0.9121	0.8508	0.6002
cosine_accuracy@150	0.9905	1.0	0.9852	0.9029	0.9418	0.8898	0.6613
cosine_accuracy@200	0.9905	1.0	0.9852	0.9417	0.9548	0.9204	0.7062
cosine_precision@1	0.6571	0.1243	0.2956	0.3495	0.4113	0.2943	0.0971
cosine_precision@20	0.5024	0.4897	0.4246	0.1733	0.0892	0.0731	0.0314
cosine_precision@50	0.308	0.3179	0.2814	0.0944	0.0418	0.0361	0.0185
cosine_precision@100	0.1863	0.1986	0.1801	0.0589	0.0229	0.0206	0.0116
cosine_precision@150	0.1322	0.1469	0.1362	0.0458	0.0159	0.0147	0.0087
cosine_precision@200	0.103	0.1179	0.1105	0.0385	0.0122	0.0116	0.0071
cosine_recall@1	0.068	0.0031	0.0111	0.0273	0.1565	0.1109	0.0329
cosine_recall@20	0.5385	0.3221	0.2614	0.1766	0.6594	0.5344	0.2091
cosine_recall@50	0.726	0.4638	0.3835	0.2393	0.7705	0.6585	0.3054
cosine_recall@100	0.8329	0.5438	0.4677	0.2863	0.8472	0.7525	0.3835
cosine_recall@150	0.8745	0.5825	0.5183	0.3287	0.8825	0.8026	0.4309
cosine_recall@200	0.9057	0.6147	0.5517	0.3631	0.9051	0.8418	0.4715
cosine_ndcg@1	0.6571	0.1243	0.2956	0.3495	0.4113	0.2943	0.0971
cosine_ndcg@20	0.6845	0.5385	0.4601	0.2468	0.5117	0.3919	0.1385
cosine_ndcg@50	0.704	0.5012	0.4229	0.2394	0.542	0.4256	0.1656
cosine_ndcg@100	0.7589	0.5147	0.4371	0.2619	0.5588	0.4462	0.1835
cosine_ndcg@150	0.7774	0.5348	0.4629	0.2787	0.5656	0.4561	0.1931
cosine_ndcg@200	0.7893	0.5505	0.4797	0.2919	0.5697	0.4632	0.2007
cosine_mrr@1	0.6571	0.1243	0.2956	0.3495	0.4113	0.2943	0.0971
cosine_mrr@20	0.8103	0.5515	0.4896	0.4485	0.4979	0.3779	0.1522
cosine_mrr@50	0.8103	0.5515	0.4909	0.4515	0.501	0.3815	0.1564
cosine_mrr@100	0.8103	0.5515	0.4911	0.4519	0.5018	0.3827	0.158
cosine_mrr@150	0.8103	0.5515	0.4912	0.4523	0.5021	0.3831	0.1585
cosine_mrr@200	0.8103	0.5515	0.4912	0.4525	0.5021	0.3832	0.1588
cosine_map@1	0.6571	0.1243	0.2956	0.3495	0.4113	0.2943	0.0971
cosine_map@20	0.5418	0.4028	0.3236	0.147	0.4264	0.3097	0.0875
cosine_map@50	0.5327	0.3422	0.2644	0.1267	0.4338	0.3174	0.093
cosine_map@100	0.5657	0.3395	0.2576	0.1326	0.436	0.3199	0.095
cosine_map@150	0.5734	0.3478	0.2669	0.1352	0.4366	0.3207	0.0957
cosine_map@200	0.5772	0.3534	0.2722	0.1368	0.4368	0.3212	0.0961
cosine_map@500	0.5814	0.3631	0.2833	0.1407	0.4373	0.3219	0.0971

Training Details

Training Datasets

full_en

full_en

• Dataset: full_en
• Size: 28,880 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 3 tokens mean: 5.0 tokens max: 10 tokens	min: 3 tokens mean: 5.01 tokens max: 13 tokens

• Samples:

  anchor	positive
  air commodore	flight lieutenant
  command and control officer	flight officer
  air commodore	command and control officer

• Loss: GISTEmbedLoss with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
    (2): Normalize()
  ), 'temperature': 0.01, 'margin_strategy': 'absolute', 'margin': 0.0}
  

full_de

full_de

• Dataset: full_de
• Size: 23,023 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 3 tokens mean: 11.05 tokens max: 45 tokens	min: 3 tokens mean: 11.43 tokens max: 45 tokens

• Samples:

  anchor	positive
  Staffelkommandantin	Kommodore
  Luftwaffenoffizierin	Luftwaffenoffizier/Luftwaffenoffizierin
  Staffelkommandantin	Luftwaffenoffizierin

• Loss: GISTEmbedLoss with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
    (2): Normalize()
  ), 'temperature': 0.01, 'margin_strategy': 'absolute', 'margin': 0.0}
  

full_es

full_es

• Dataset: full_es
• Size: 20,724 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 3 tokens mean: 12.95 tokens max: 50 tokens	min: 3 tokens mean: 12.57 tokens max: 50 tokens

• Samples:

  anchor	positive
  jefe de escuadrón	instructor
  comandante de aeronave	instructor de simulador
  instructor	oficial del Ejército del Aire

• Loss: GISTEmbedLoss with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
    (2): Normalize()
  ), 'temperature': 0.01, 'margin_strategy': 'absolute', 'margin': 0.0}
  

full_zh

full_zh

• Dataset: full_zh
• Size: 30,401 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 4 tokens mean: 8.36 tokens max: 20 tokens	min: 4 tokens mean: 8.95 tokens max: 27 tokens

• Samples:

  anchor	positive
  技术总监	技术和运营总监
  技术总监	技术主管
  技术总监	技术艺术总监

• Loss: GISTEmbedLoss with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
    (2): Normalize()
  ), 'temperature': 0.01, 'margin_strategy': 'absolute', 'margin': 0.0}
  

mix

mix

• Dataset: mix
• Size: 21,760 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 2 tokens mean: 5.65 tokens max: 14 tokens	min: 2 tokens mean: 10.08 tokens max: 30 tokens

• Samples:

  anchor	positive
  technical manager	Technischer Direktor für Bühne, Film und Fernsehen
  head of technical	directora técnica
  head of technical department	技术艺术总监

• Loss: GISTEmbedLoss with these parameters:

  {'guide': SentenceTransformer(
    (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 
    (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
    (2): Normalize()
  ), 'temperature': 0.01, 'margin_strategy': 'absolute', 'margin': 0.0}
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 128
• per_device_eval_batch_size: 128
• gradient_accumulation_steps: 2
• num_train_epochs: 5
• warmup_ratio: 0.05
• log_on_each_node: False
• fp16: True
• dataloader_num_workers: 4
• ddp_find_unused_parameters: True
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 128
• per_device_eval_batch_size: 128
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 2
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 5
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.05
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: False
• logging_nan_inf_filter: True
• save_safetensors: True
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• no_cuda: False
• use_cpu: False
• use_mps_device: False
• seed: 42
• data_seed: None
• jit_mode_eval: False
• use_ipex: False
• bf16: False
• fp16: True
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: 0
• ddp_backend: None
• tpu_num_cores: None
• tpu_metrics_debug: False
• debug: []
• dataloader_drop_last: True
• dataloader_num_workers: 4
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: False
• ignore_data_skip: False
• fsdp: []
• fsdp_min_num_params: 0
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• tp_size: 0
• fsdp_transformer_layer_cls_to_wrap: None
• accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• ddp_find_unused_parameters: True
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• skip_memory_metrics: True
• use_legacy_prediction_loop: False
• push_to_hub: False
• resume_from_checkpoint: None
• hub_model_id: None
• hub_strategy: every_save
• hub_private_repo: None
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• include_for_metrics: []
• eval_do_concat_batches: True
• fp16_backend: auto
• push_to_hub_model_id: None
• push_to_hub_organization: None
• mp_parameters:
• auto_find_batch_size: False
• full_determinism: False
• torchdynamo: None
• ray_scope: last
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• include_tokens_per_second: False
• include_num_input_tokens_seen: False
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	full_en_cosine_ndcg@200	full_es_cosine_ndcg@200	full_de_cosine_ndcg@200	full_zh_cosine_ndcg@200	mix_es_cosine_ndcg@200	mix_de_cosine_ndcg@200	mix_zh_cosine_ndcg@200
-1	-1	-	0.7322	0.4690	0.3853	0.2723	0.3209	0.2244	0.0919
0.0021	1	23.8878	-	-	-	-	-	-	-
0.2058	100	7.2098	-	-	-	-	-	-	-
0.4115	200	4.2635	0.7800	0.5132	0.4268	0.2798	0.4372	0.2996	0.1447
0.6173	300	4.1931	-	-	-	-	-	-	-
0.8230	400	3.73	0.7863	0.5274	0.4451	0.2805	0.4762	0.3455	0.1648
1.0309	500	3.3569	-	-	-	-	-	-	-
1.2366	600	3.6464	0.7868	0.5372	0.4540	0.2813	0.5063	0.3794	0.1755
1.4424	700	3.0772	-	-	-	-	-	-	-
1.6481	800	3.114	0.7906	0.5391	0.4576	0.2832	0.5221	0.4047	0.1779
1.8539	900	2.9246	-	-	-	-	-	-	-
2.0617	1000	2.7479	0.7873	0.5423	0.4631	0.2871	0.5323	0.4143	0.1843
2.2675	1100	3.049	-	-	-	-	-	-	-
2.4733	1200	2.6137	0.7878	0.5418	0.4685	0.2870	0.5470	0.4339	0.1932
2.6790	1300	2.8607	-	-	-	-	-	-	-
2.8848	1400	2.7071	0.7889	0.5465	0.4714	0.2891	0.5504	0.4362	0.1944
3.0926	1500	2.7012	-	-	-	-	-	-	-
3.2984	1600	2.7423	0.7882	0.5471	0.4748	0.2868	0.5542	0.4454	0.1976
3.5041	1700	2.5316	-	-	-	-	-	-	-
3.7099	1800	2.6344	0.7900	0.5498	0.4763	0.2857	0.5639	0.4552	0.1954
3.9156	1900	2.4983	-	-	-	-	-	-	-
4.1235	2000	2.5423	0.7894	0.5499	0.4786	0.2870	0.5644	0.4576	0.1974
4.3292	2100	2.5674	-	-	-	-	-	-	-
4.5350	2200	2.6237	0.7899	0.5502	0.4802	0.2843	0.5674	0.4607	0.1993
4.7407	2300	2.3776	-	-	-	-	-	-	-
4.9465	2400	2.1116	0.7893	0.5505	0.4797	0.2919	0.5697	0.4632	0.2007

Framework Versions

• Python: 3.11.11
• Sentence Transformers: 4.1.0
• Transformers: 4.51.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.6.0
• Datasets: 3.5.0
• Tokenizers: 0.21.1

Citation

BibTeX

Sentence Transformers

@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",
}

GISTEmbedLoss

@misc{solatorio2024gistembed,
    title={GISTEmbed: Guided In-sample Selection of Training Negatives for Text Embedding Fine-tuning},
    author={Aivin V. Solatorio},
    year={2024},
    eprint={2402.16829},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}