cometadata/jina-reranker-v2-multilingual-affiliations-comet-affilgood-training-mix

This is a Cross Encoder model finetuned from jinaai/jina-reranker-v2-base-multilingual using the sentence-transformers library. It computes scores for pairs of texts, which can be used for text reranking and semantic search.

Model Details

Model Description

• Model Type: Cross Encoder
• Base model: jinaai/jina-reranker-v2-base-multilingual
• Maximum Sequence Length: 1024 tokens
• Number of Output Labels: 1 label
• Language: multilingual
• License: cc-by-nc-4.0

Model Sources

• Documentation: Sentence Transformers Documentation
• Documentation: Cross Encoder Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Cross Encoders on Hugging Face

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("cometadata/jina-reranker-v2-multilingual-affiliations-v4")
# Get scores for pairs of texts
pairs = [
    ['Université Toulouse', 'a  Université de Toulouse, Mines Albi, CNRS, Centre RAPSODEE ,  Albi ,  France'],
    ['Université Toulouse', 'National Polytechnic Institute of Toulouse'],
    ['School of Fundamental Science and Technology, Keio University 1 , 3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan', 'Center for Supercentenarian Research, Keio University, Tokyo, Japan'],
    ['School of Fundamental Science and Technology, Keio University 1 , 3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan', 'g    Toin Human Science and Technology Center, Department of Materials Science and Technology, Toin University of Yokohama, 1614 Kurogane-cho, Aoba-ku, Yokohama 225, Japan'],
    ['Division of Pulmonary and Critical Care Medicine, University of North Carolina School of Medicine, Chapel Hill, North Carolina', 'Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, 101 Manning Drive, CB# 7295, Chapel Hill, NC 27599, USA'],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'Université Toulouse',
    [
        'a  Université de Toulouse, Mines Albi, CNRS, Centre RAPSODEE ,  Albi ,  France',
        'National Polytechnic Institute of Toulouse',
        'Center for Supercentenarian Research, Keio University, Tokyo, Japan',
        'g    Toin Human Science and Technology Center, Department of Materials Science and Technology, Toin University of Yokohama, 1614 Kurogane-cho, Aoba-ku, Yokohama 225, Japan',
        'Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, 101 Manning Drive, CB# 7295, Chapel Hill, NC 27599, USA',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Reranking

• Dataset: affiliation-val
• Evaluated with CrossEncoderRerankingEvaluator with these parameters:

  {
      "at_k": 10,
      "always_rerank_positives": true
  }
  

Metric	Value
map	0.9307 (-0.0693)
mrr@10	0.9307 (-0.0693)
ndcg@10	0.9502 (-0.0498)

Training Details

Training Dataset

Unnamed Dataset

• Size: 16,862 training samples
• Columns: query, document, and label
• Approximate statistics based on the first 1000 samples:

  	query	document	label
  type	string	string	int
  details	min: 6 characters mean: 95.73 characters max: 505 characters	min: 8 characters mean: 92.11 characters max: 393 characters	0: ~50.00% 1: ~50.00%

• Samples:

  query	document	label
  Nanjing University of Science and Technology,Computer Science and Engineering,Nanjing,China	Nanjing University of Science And Technology, China	1
  Nanjing University of Science and Technology,Computer Science and Engineering,Nanjing,China	Nanjing university of finance & economics, China.	0
  University of Bonn, Bonn, Germany	Department of Geophysics, University of Bonn, 53115 Bonn, Germany	1

• Loss: BinaryCrossEntropyLoss with these parameters:

  {
      "activation_fn": "torch.nn.modules.linear.Identity",
      "pos_weight": null
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 808 evaluation samples
• Columns: query, document, and label
• Approximate statistics based on the first 808 samples:

  	query	document	label
  type	string	string	int
  details	min: 14 characters mean: 80.47 characters max: 394 characters	min: 15 characters mean: 109.87 characters max: 500 characters	0: ~50.00% 1: ~50.00%

• Samples:

  query	document	label
  Université Toulouse	a Université de Toulouse, Mines Albi, CNRS, Centre RAPSODEE , Albi , France	1
  Université Toulouse	National Polytechnic Institute of Toulouse	0
  School of Fundamental Science and Technology, Keio University 1 , 3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan	Center for Supercentenarian Research, Keio University, Tokyo, Japan	1

• Loss: BinaryCrossEntropyLoss with these parameters:

  {
      "activation_fn": "torch.nn.modules.linear.Identity",
      "pos_weight": null
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• learning_rate: 2e-05
• num_train_epochs: 2
• warmup_ratio: 0.1
• bf16: True
• load_best_model_at_end: True
• hub_model_id: cometadata/jina-reranker-v2-multilingual-affiliations-v4

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 32
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 2e-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: 2
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.1
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• 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
• bf16: True
• fp16: False
• 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: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: True
• 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}
• 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}
• parallelism_config: None
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• adafactor: False
• group_by_length: False
• length_column_name: length
• project: huggingface
• trackio_space_id: trackio
• ddp_find_unused_parameters: None
• 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: cometadata/jina-reranker-v2-multilingual-affiliations-v4
• hub_strategy: every_save
• hub_private_repo: None
• hub_always_push: False
• hub_revision: None
• 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: no
• neftune_noise_alpha: None
• optim_target_modules: None
• batch_eval_metrics: False
• eval_on_start: False
• use_liger_kernel: False
• liger_kernel_config: None
• eval_use_gather_object: False
• average_tokens_across_devices: True
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss	affiliation-val_ndcg@10
-1	-1	-	-	0.8812 (-0.1188)
0.0019	1	0.577	-	-
0.1898	100	0.5546	-	-
0.3795	200	0.3925	-	-
0.5693	300	0.3369	-	-
0.7590	400	0.3175	-	-
0.9488	500	0.3233	0.5399	0.9502 (-0.0498)
1.1385	600	0.2847	-	-
1.3283	700	0.2864	-	-
1.5180	800	0.3	-	-
1.7078	900	0.2782	-	-
1.8975	1000	0.2783	0.528	0.9502 (-0.0498)
-1	-1	-	-	0.9502 (-0.0498)

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.2.0
• Transformers: 4.57.3
• PyTorch: 2.9.1+cu128
• Accelerate: 1.12.0
• Datasets: 4.4.2
• Tokenizers: 0.22.1

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",
}