SentenceTransformer based on sentence-transformers/sentence-t5-base

This is a sentence-transformers model finetuned from sentence-transformers/sentence-t5-base. 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: sentence-transformers/sentence-t5-base
• 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': 'T5EncoderModel'})
  (1): Pooling({'embedding_dimension': 768, 'pooling_mode': 'mean', 'include_prompt': True})
  (2): Dense({'in_features': 768, 'out_features': 768, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity', 'module_input_name': 'sentence_embedding', 'module_output_name': 'sentence_embedding'})
  (3): 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("RinKana/sentence-t5-base-iran-us-news-embedder")
# Run inference
sentences = [
    'Strait of Hormuz crisis',
    "Commentary: Strait of Hormuz is trapped between geopolitics and legal vacuum | [Editor's Note: Follow Khaleej Times' live blog for the latest regional developments with the US-Israel-Iran ceasefire now in effect.]\n\nThe author is president of UAE Safety and Emergency Security Association and professor of sustainable development at the American University of Sharjah. He is former UAE Minister of Climate Change and Environment and former Minister of Infrastructure. He authored two books on climate change that influenced academic and international reforms.\n\nWhat pushed me to revisit the issue of the double siege on the Strait of Hormuz was not only the gravity of the current geopolitical moment but also the extent of the confusion surrounding the strait’s legal status.\n\nIt has become increasingly common to hear analysts cite provisions of the United Nations Convention on the Law of the Sea (UNCLOS) as if they automatically apply to Hormuz, despite the fact that neither Iran nor the United States is a party to the convention. This alone strips those provisions of any binding force in this context.\n\nWorse still, some commentators invoke treaties that simply do not exist in maritime law, such as the so-called Gomika Treaty, or rely on texts unrelated to international straits in an attempt to give political or military positions a veneer of legal legitimacy.\n\nStay up to date with the latest news. Follow KT on WhatsApp Channels.\n\nThis confusion between law and reality, between texts and facts, creates a misleading public picture and makes it necessary to present a precise reading grounded in what actually exists: A legal vacuum, a double siege, a maritime organisation with no enforcement tools, and a global economy suspended between fear and anticipation.\n\nIn a historical moment where geography intersects with power and law dissolves into emptiness, the world finds itself facing an unprecedented maritime scene: the Strait of Hormuz — the artery through which a fifth of global energy flows — has turned into a theatre of dual pressure from inside and outside.\n\nMaritime choke point\n\nAs regional and international powers exchange signals of escalation, global markets stand on edge, and the international economy lives in a state of fear and vigilance, wary that this maritime choke point could trigger an uncontrollable global energy crisis.\n\nNearly 20 million barrels per day of oil and petroleum liquids pass through the strait — around 20 per cent of global consumption — in addition to 20 per cent of global LNG trade. It is the artery on which Asia, led by China, relies to feed its heavy and petrochemical industries and to maintain energy price stability.\n\nWith Iran’s internal pressure and the broader American pressure from the outside, the world faces an unprecedented situation: The most critical energy strait on earth is caught between two states that are not parties to the Law of the Sea, under the watch of a maritime organisation with no enforcement authority.\n\nThe absence of Iranian and American commitment to UNCLOS means that the principle of “transit passage”, which protects navigation in international straits, cannot be invoked legally. Nor is it possible to resort to the International Tribunal for the Law of the Sea, while the International Court of Justice requires the consent of both parties, which is an unlikely scenario. This leaves customary international law – far less defined – as the only reference, opening the door to conflicting interpretations and turning the strait into a space governed by power rather than law.\n\nWithin this legal vacuum, the International Maritime Organization (IMO) faces a crisis beyond its mandate. It has no naval forces, no sanctions regime, and no ability to compel a state to open a strait or prevent another from imposing a blockade. Its tools are limited to navigational warnings, technical recommendations, condemnation of attacks on ships, and reminders of obligations under safety, pollution, and ship‑security conventions.\n\nImmediate and severe impact\n\nThe real penalty does not come from the organisation itself but from the reaction of the global maritime system: Insurers raise premiums or halt coverage, ports tighten procedures on certain flags, and shipping companies reroute their fleets — all of which translate into additional costs for states that ignore international standards. It is a form of soft pressure, but it is no substitute for the hard enforcement power held by the Security Council or military coalitions.\n\nThe economic impact of closure is immediate and severe. The disappearance of 8–10 million barrels per day of accessible supply — due to the limited alternatives offered by pipelines — triggers sharp price spikes, supply-chain disruptions, and soaring shipping and insurance costs.\n\nMarkets react not only to actual volumes but also to the fear of the unknown: How long will the closure last? Could the conflict expand? This is where the real nervousness emerges: violent swings in equity markets, pressure on the currencies of energy‑importing states, rising air and sea transport costs, and heightened stress across economic confidence indicators.\n\nIt is also important to correct the widespread belief that rising oil and gas prices benefit exporting states. In reality, Gulf exporters face direct and immediate losses because the problem is not the price — it is the inability to export.\n\nAs tankers stop moving, storage fills quickly, forcing states into involuntary production cuts, meaning large daily revenue losses even if global prices rise. Shipping and insurance costs surge, risk premiums multiply, and some vessels refuse to enter the region altogether. Condensates are even more sensitive than crude, as they cannot be stored in large quantities, and halting their export disrupts gas‑processing units, affecting electricity generation, domestic gas supply, and LNG output.\n\nThe result is clear – higher prices do not compensate for lost volumes, and exporting states face a mix of declining revenues, rising costs, domestic market disruptions, and eroding international confidence in supply stability.\n\nStrait of chronic tension\n\nAlthough the situation appears unprecedented, history offers parallels showing that maritime straits have always been strategic vulnerabilities. The closure of the Suez Canal between 1967 and 1975 — the longest shutdown of a global waterway — forced ships around the Cape of Good Hope, raised shipping costs, and left the “Yellow Fleet” stranded for eight years.\n\nThe Bab al‑Mandeb Strait saw partial closure in 1973, affecting oil flows to Europe. During World War I, the Ottoman Empire used the Bosporus and Dardanelles as pressure points against Russia. Even Hormuz itself has known similar episodes: Portuguese control in the 16th century and British pressure on Iran in the mid‑20th century. These precedents are not identical to today’s situation, but they confirm that straits have always been theatres of conflict and that international law alone is insufficient when geography intersects with power.\n\nWhat makes the current moment unique is the convergence of three rare elements: A double siege from inside and outside, an almost complete absence of a binding legal framework governing international straits, and a global dependence on Hormuz far greater than that of any other strait in history.\n\nUnder this equation, markets will remain in a state of chronic tension, and China — along with the rest of Asia — will continue searching for a new formula to reduce its economic exposure to a single narrow corridor, perpetually open to escalation.",
    'We need more plumbers and fewer lawyers in AI age, says BlackRock boss | Society should respect plumbing and electrician careers as artificial intelligence (AI) replaces some office jobs, the boss of US financial giant BlackRock has told the BBC.\n\nLarry Fink, who leads the world\'s largest asset manager, says the US "overdid" the focus on young people going to university and idolising jobs careers in banking while undervaluing skilled trades.\n\nIn a wide-ranging exclusive interview, he also warned if the price of oil hits $150 (£122) a barrel it will trigger a global recession.\n\nBlackRock is a financial colossus, controlling assets worth $14 trillion, and is one of the biggest investors in many of the world\'s largest companies.\n\nIts size and spread gives Fink, who is one of the eight co-founders of the business, which started in 1988, a unique insight into the health of the global economy.\n\nEarlier this week, in his annual letter to shareholders, external, Fink said the AI boom would create an enormous amount of jobs "related to electricians and welders and plumbers".\n\nIn contrast, there might not be as much demand for some office jobs and this could lead to a rethink about what roles are needed as "society is changing and evolving".\n\nHe said this meant we needed to change our perception of skilled trades.\n\nHe said the average plumber had been portrayed on television as being overweight and having their pants hanging below their waistline, while investment bankers are idolised in drama series like Industry.\n\n"I think what we did was wrong," he said. "We really put judgment on so many jobs and so many people who probably should not have gone into banking or media or law, probably should have been a great worker with their hands, and we need to now rebalance that approach."\n\nIn the US, he says, after World War Two "we built the foundation of education, and we said to all the young people, go to college, go to college, go to college. And we probably overdid it".\n\n"We need to balance that out, and we need to be proud that... a career can be just as strong in these fields of plumbing and electricians."',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[nan, nan, nan],
#         [nan, nan, nan],
#         [nan, nan, nan]], dtype=torch.float16)

Evaluation

Metrics

Information Retrieval

• Dataset: val-ir
• Evaluated with InformationRetrievalEvaluator

Training Details

Training Dataset

Unnamed Dataset

• Size: 2,057 training samples

• Columns: sentence_0 and sentence_1

• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1
  type	string	string
  details	min: 5 tokensmean: 6.06 tokensmax: 9 tokens	min: 53 tokensmean: 230.92 tokensmax: 256 tokens

• Samples:

  sentence_0	sentence_1
  Twelve-Day War Iran	Watch: Iran's World Cup team arrives in Mexico amid US visa row | The Iranian football sqaud has arrived in Tijuana, Mexico, which will serve as its base for the 2026 World Cup amid a row over US visas.The side's original plan to stay in Tucson, Arizona was changed after visas were denied for several members of staff, including the head of Iran's football federation, Mehdi Taj.All of Iran's group games are in the US, but players will have to fly in and out of the country on each match day.It comes as the US-Israeli war with Iran continues, making this the first World Cup to see a host nation receive the team of a country it is at war with.
  Iran Israel war 2025	100 days of the US-Israeli war of aggression against Iran

  Lebanese security officers gather at the site where an Israeli airstrike hit a building in Dahiyeh, Beirut's southern suburb, Lebanon, Sunday, June 7, 2026. [AP Photo/Hassan Ammar]

  One hundred days ago, on February 28, the United States and Israel launched an illegal war of aggression against Iran. The war is being waged by the world’s most powerful imperialist powers against a historically oppressed nation.

  The resistance of the Iranian people, notwithstanding the reactionary character of the clerical regime, is politically legitimate and of a heroic character. The working class internationally must defend Iran unconditionally against imperialist subjugation.

  The “negotiations” currently being carried out by the Trump administration at gunpoint are a fraud. In an interview this weekend, Trump declared that if Iran does not accept his demands, “I’m going to blow the hell out of them.” Even if ... | | Iran Trump 2026 | Trump Calls Off Iran Strikes with Deal in Reach: “We Should Be Getting It Done In The Next Few Days.” | Bloomberg’s Balance of Power covered the story on Thursday, June 11, that President Trump announced he was calling off the planned U.S. military strikes against Iran, with a memorandum of understanding potentially being signed within days. Markets responded immediately: the S&P 500 closed up 1.8%, and the Nasdaq rallied close to 3% on the de-escalation signal, while oil prices fell 4-5% intraday, with WTI crude cited on Friday, June 12, at below $85 per barrel.

  The president framed the outcome in decisive terms: “We made a great settlement of the war with Iran, and we are going to be submitting the finalization of documents. We should be getting it done in the next few days.“ He acknowledged loose ends but defended the structure of the agreement, saying, “It is a very strong memorandum of understanding. It’s a little conceptual, but it is something that’s going to get done.”

  Why Ma... |

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "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 Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 2
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

• do_predict: False
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• gradient_accumulation_steps: 1
• 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: 2
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_ratio: None
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• enable_jit_checkpoint: False
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• use_cpu: False
• seed: 42
• data_seed: None
• bf16: False
• fp16: False
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: -1
• ddp_backend: None
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: False
• ignore_data_skip: False
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
• 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
• 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
• 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_for_metrics: []
• eval_do_concat_batches: True
• auto_find_batch_size: False
• full_determinism: False
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• 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
• use_cache: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Training Time

• Training: 1.4 minutes

Framework Versions

• Python: 3.12.13
• Sentence Transformers: 5.4.1
• Transformers: 5.0.0
• PyTorch: 2.10.0+cu128
• Accelerate: 1.13.0
• Datasets: 5.0.0
• Tokenizers: 0.22.2

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",
}

MultipleNegativesRankingLoss

@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},
}