This is a sentence-transformers model finetuned from jinaai/jina-embeddings-v5-text-nano-retrieval. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for retrieval.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: jinaai/jina-embeddings-v5-text-nano-retrieval
Maximum Sequence Length: 256 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Supported Modality: Text

Model Sources

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

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': 'EuroBertModel'})
  (1): Pooling({'embedding_dimension': 768, 'pooling_mode': 'lasttoken', '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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("MinhPhuc0804/jina-docling-checkthat-task1-v1")
# Run inference
queries = [
    "query: @MikeHudema Check out this review article for further discussion on why fossil fuels will keep supplying the bulk of global energy and why electric vehicles aren't exactly eco‑friendly",
]
documents = [
    'passage: title: Is it the end of combustion and engine combustion research? Should it be?\nabstract: The dominant narrative in the affluent west is that climate change poses an "existential threat" and very rapid cuts in greenhouse gas (GHG) emissions and hence fossil fuel use are needed to avoid it. Simultaneously oil, gas, coal, aviation, steel and cement industries and livestock farming have to be largely shut down to eliminate GHG. This review argues that globally all this will not happen by 2050, let alone 2030 because the scale of the problem is too large. Transport is particularly difficult to decarbonize and current policies focusing entirely on battery electric vehicles will not and must not succeed. GHG levels are unlikely to come down significantly in the next several decades and even if they did, extreme weather events will not disappear. It is better to recognize such realities and make societies more resilient to the effects of climate change. Humanity will have to adapt to any further warming as it has successfully done with the previous warming of about 1.1 C over the past century. Combustion research, particularly of fossil fuels and in internal combustion engines is currently seen as unnecessary in many countries.',
    'passage: title: SARS-CoV-2 infection confers greater immunity than shots\nabstract: Study from Israel, the largest of its kind, also finds infection combined with a single jab is highly protective',
    'passage: title: Neurologic Involvement in Children and Adolescents Hospitalized in the United States for COVID-19 or Multisystem Inflammatory Syndrome\nabstract: Coronavirus disease 2019 (COVID-19) affects the nervous system in adult patients. The spectrum of neurologic involvement in children and adolescents is unclear.To understand the range and severity of neurologic involvement among children and adolescents associated with COVID-19.Case series of patients (age <21 years) hospitalized between March 15, 2020, and December 15, 2020, with positive severe acute respiratory syndrome coronavirus 2 test result (reverse transcriptase-polymerase chain reaction and/or antibody) at 61 US hospitals in the Overcoming COVID-19 public health registry, including 616 (36%) meeting criteria for multisystem inflammatory syndrome in children. Patients with neurologic involvement had acute neurologic signs, symptoms, or diseases on presentation or during hospitalization.',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 768] [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[ 0.5678, -0.0192, -0.0984]])

Evaluation

Metrics

Information Retrieval

Dataset: 10-percent-dev-split
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.5299
cosine_accuracy@3	0.7273
cosine_accuracy@5	0.787
cosine_accuracy@10	0.8447
cosine_precision@1	0.5299
cosine_precision@3	0.2424
cosine_precision@5	0.1574
cosine_precision@10	0.0845
cosine_recall@1	0.5299
cosine_recall@3	0.7273
cosine_recall@5	0.787
cosine_recall@10	0.8447
cosine_ndcg@10	0.6898
cosine_mrr@10	0.6399
cosine_map@100	0.6446

Training Details

Training Dataset

Unnamed Dataset

Size: 17,319 training samples
Columns: query and positive
Approximate statistics based on the first 1000 samples:
query positive
type string string
details
min: 21 tokens
mean: 53.57 tokens
max: 144 tokens
min: 27 tokens
mean: 183.22 tokens
max: 256 tokens

Samples:

query	positive
`query: @user Abgesehen von den bekannten Nebenwirkungen wie Thrombosen und Neurologischen Schäden gibt es zum Beispiel jedoch begutachtete Anhaltspunkte, dass das Protein die DNA schädigt. Da habe ich keine Lust darauf.`	`passage: its DNA damage repair by impeding key DNA repair protein BRCA1 and 53BP1 recruitment to the damage site.[object Object][object Object]title: RETRACTED: SARS–CoV–2 Spike Impairs DNA Damage Repair and Inhibits V(D)J Recombination In Vitro[object Object]Our findings reveal a potential molecular mechanism by which the spike protein might impede adaptive immunity and underscore the potential side effects of full-length spike-based vaccines.`
`query: de plus le Japon emploie également les échantillons salivaires ; ci le lien vers l’article de Yale, je vous prie de vérifier les points que je présente et de corriger. 3/`	passage: ngeal and saliva samples from confirmed COVID-19 patients and self-collected samples from healthcare workers on COVID-19 wards.[object Object][object Object]title: Saliva is more sensitive for SARS-CoV-2 detection in COVID-19 patients than nasopharyngeal swabs[object Object]When we compared SARS-CoV-2 detection from patient-matched nasopharyngeal and saliva samples, we found that saliva yielded greater detection sensitivity and consistency throughout the course of infection. Furthermore, we report less variability in self-sample collection of saliva. Taken together, our findings demonstrate that saliva is a viable and more sensitive alternative to nasopharyngeal swabs and could enable at-home self-administered sample collection for accurate large-scale SARS-CoV-2 testing.
`query: Are psychedelics the road to comprehension? Interpersonal Dynamics in Ayahuasca circles of Palestinians and Israelis`	`passage: title: Relational Processes in Ayahuasca Groups of Palestinians and Israelis`
abstract: Psychedelics are used in many group contexts. However, most phenomenological research on psychedelics is focused on personal experiences. This paper presents a phenomenological investigation centered on intersubjective and intercultural relational processes, exploring how an intercultural context affects both the group and individual process. Through 31 in-depth interviews, ceremonies in which Palestinians and Israelis drink ayahuasca together have been investigated. The overarching question guiding this inquiry was how psychedelics might contribute to processes of peacebuilding, and in particular how an intercultural context, embedded in a protracted conflict, would affect the group's psychedelic process in a relational sense. Analysis of the interviews was based on grounded theory. Three relational themes about multilocal participatory events which occurred during ayahuasca rituals have em...

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim", 4 "gather_across_devices": false, 5 "directions": [ 6 "query_to_doc" 7 ], 8 "partition_mode": "joint", 9 "hardness_mode": null, 10 "hardness_strength": 0.0 11}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 32
learning_rate: 8e-06
num_train_epochs: 10
warmup_steps: 542
bf16: True
dataloader_drop_last: True
load_best_model_at_end: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
prediction_loss_only: True
per_device_train_batch_size: 32
per_device_eval_batch_size: 8
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: 8e-06
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 10
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
warmup_steps: 542
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
use_ipex: 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: True
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
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: None
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: False
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: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	10-percent-dev-split_cosine_ndcg@10
0.9242	500	0.6731	-
1.0018	542	-	0.6814
1.8484	1000	0.3426	-
2.0037	1084	-	0.6826
2.7726	1500	0.2901	-
3.0055	1626	-	0.6871
3.6969	2000	0.2571	-
4.0074	2168	-	0.6886
4.6211	2500	0.2435	-
5.0092	2710	-	0.6887
5.5453	3000	0.2322	-
6.0111	3252	-	0.6896
6.4695	3500	0.2207	-
7.0129	3794	-	0.6902
7.3937	4000	0.2199	-
8.0148	4336	-	0.6898
8.3179	4500	0.2194	-
9.0166	4878	-	0.6898
9.2421	5000	0.2129	-

The bold row denotes the saved checkpoint.

Training Time

Training: 16.2 minutes

Framework Versions

Python: 3.12.6
Sentence Transformers: 5.4.1
Transformers: 4.56.0
PyTorch: 2.8.0+cu129
Accelerate: 1.10.1
Datasets: 4.8.4
Tokenizers: 0.22.0

Citation

BibTeX

Sentence Transformers

bibtex

@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",
}

MultipleNegativesRankingLoss

bibtex

@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},
}

	query	positive
type	string	string
details	min: 21 tokens mean: 53.57 tokens max: 144 tokens	min: 27 tokens mean: 183.22 tokens max: 256 tokens