IrvinTopi
/

marginmse-webfaq

@@ -1,805 +1,87 @@
 ---
 tags:
-- sentence-transformers
 - sentence-similarity
-- feature-extraction
-- dense
-- generated_from_trainer
-- dataset_size:12753278
-- loss:MarginMSELoss
-base_model: PaDaS-Lab/xlm-roberta-base-msmarco
-widget:
-- source_sentence: Hur gammal måste jag vara för att betta på Melodifestivalen?
-  sentences:
-  - För att registrera ett spelkonto och betta online måste du vara över 18 år i Sverige.
-  - Ja, för att delta i PrisPicks plattformen måste man vara minst 18 år gammal. Denna
-    åldersgräns kan vara högre i vissa jurisdiktioner, så det rekommenderas att potentiella
-    användare kontrollerar de specifika ålderskraven för deras plats.
-  - Bu sorunun cevabı barındırdığınız sistemin büyüklüğüne, trafiğine ve optimizasyonuna
-    göre değişmektedir. Çok iyi optimize edilmiş bir sisteminiz (script vb.) ve gelen
-    trafiği optimum düzeyde karşılayacak donanımınız varsa minimum düzeyde bir sunucu
-    yeterli olacaktır. Ancak iyi optimize edilmemiş bir sistem ve sunucu için farklı
-    alternatifler aramanız gerekebilir. En iyi sunucu nedir sorusunun cevabı, sisteminize
-    ve trafiğinize göre değişebilir.
-- source_sentence: क्या लॉजिकल रीजनिंग यूजीसी नेट परीक्षा का हिस्सा है?
-  sentences:
-  - हां, लॉजिकल रीजनिंग यूजीसी नेट परीक्षा का हिस्सा है।
-  - 'Per il momento, non sono ancora entrate in vigore sul massimale minimo per le
-    polizze rc professionale medici. Teniamo conto però di una cosa: se si lavora
-    (e si è lavorato nei dieci anni precedenti) esclusivamente come dipendenti o specializzandi
-    presso l’SSN, dobbiamo sapere che la rivalsa massima dell’SSN sarà plafonata al
-    triplo del reddito annuo lordo del medico.
-    Se invece si lavora in libera professione, non c’è alcun limite. Consigliamo comunque
-    di scegliere massimali non inferiori al milione di euro.'
-  - यूजीसी नेट की परीक्षा साल में दो बार आयोजित की जाती है। प्रथम परीक्षा जून में
-    और द्वितीय परीक्षा दिसंबर महीने में नेशनल टेस्टिंग एजेंसी द्वारा आयोजित की जाती
-    है।
-- source_sentence: Car Accident Lawyer in Denver, CO
-  sentences:
-  - A Vinsa Telêmaco Borba é uma empresa com sede no município de e fica localizada
-    na Al. Washington Luiz, 490 – Alto das Oliveiras – Telêmaco Borba – PR.
-  - When you are in need of a skilled car accident lawyer, a lawyer in the Denver,
-    CO area, don’t wait to talk to a lawyer from The Law Offices of Cliff Enten. With
-    years of legal experience, they have provided excellent results for their clients
-    in many types of personal injury areas. They are knowledgeable about the tactics
-    the other party will use to get the highest possible compensation amount. For
-    more information set up an appointment with them now.
-  - After seeking immediate medical attention for your injuries, you may have your
-    case evaluated by a professional Denver personal injury lawyer at Mintz Law Firm
-    and obtain expert legal advice. Regardless of the type of accident or event and
-    the apparent extent of your injuries, a lawyer can help you pursue the compensation
-    owed to you because the negligent parties are responsible for the damage. After
-    any car accident, slip and fall, or animal bite, don’t delay contacting our legal
-    team to help you strategize and take care of your case.
-- source_sentence: Wie wird Blood Suckers gespielt?
-  sentences:
-  - Ja, die erste Version von Blood Suckers war so erfolgreich, dass es mittlerweile
-    sogar eine zweite Variante gibt. Der Automat Blood Suckers 2 wurde 2017 veröffentlicht.
-    Hier gibt es das gleiche Motto und 25 Gewinnlinien. Zwar wird die Geschichte diesmal
-    weiter erzählt.
-  - La bozza del prodotto ordinato ti arriverà entro 24 ore dal tuo acquisto tramite
-    e-mail. Se non riesci a visualizzarla ti consigliamo di controllare nello spam
-    della tua posta elettronica. Non ti preoccupare puoi trovare la tua bozza anche
-    nella sezione i miei ordini. In corrispondenza del prodotto acquistato troverai
-    un pulsante
-  - Sie spielen auf einem Raster von 5 Walzen und drei Reihen mit 25 Gewinnlinien.
-    Beim Blood Suckers Spiel gewinnen Sie von links nach rechts.
-- source_sentence: Kiedy rozpoczyna się kurs?
-  sentences:
-  - 'Para convertirte en representante de ventas debes entender lo que son los productos
-    de tecnología o del sector turístico para vender de manera efectiva. Utiliza tus
-    conexiones y conocimiento, obtén clientes y genera ingresos. Debes familiarizarte
-    con los procesos de ventas y marketing de productos de tecnología y/o de viaje,
-    y estar activamente preparado para promover la plataforma moonstride.
-    Para unirte a nuestro programa, tu negocio debe ser legítimo y estar bien establecido.
-    moonstride realizará un proceso de selección para asegurar la calidad de nuestros
-    representantes'
-  - Nowe kursy grupowe zaczynają się w połowie lutego. Dokładna data jest przekazana
-    kursantom tydzień przed rozpoczęciem kursu.
-  - Kurs startuje dla Ciebie w momencie, kiedy się na niego zapiszesz. Chwilę później
-    otrzymasz pierwszą lekcję oraz obiecany do niej BONUS w postaci e-booka „100+
-    nieoczywistych pomysłów na e-sklep”. Kolejne lekcje będziemy wysyłać co 3 dni.
-pipeline_tag: sentence-similarity
-library_name: sentence-transformers
 ---
-# SentenceTransformer based on PaDaS-Lab/xlm-roberta-base-msmarco
-This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [PaDaS-Lab/xlm-roberta-base-msmarco](https://huggingface.co/PaDaS-Lab/xlm-roberta-base-msmarco). It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
-## Model Details
-### Model Description
-- **Model Type:** Sentence Transformer
-- **Base model:** [PaDaS-Lab/xlm-roberta-base-msmarco](https://huggingface.co/PaDaS-Lab/xlm-roberta-base-msmarco) <!-- at revision cd02f4c38b71baa0dc6b3fcdd86a3b6bd407ef55 -->
-- **Maximum Sequence Length:** 512 tokens
-- **Output Dimensionality:** 768 dimensions
-- **Similarity Function:** Cosine Similarity
-<!-- - **Training Dataset:** Unknown -->
-<!-- - **Language:** Unknown -->
-<!-- - **License:** Unknown -->
-### Model Sources
-- **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
-- **Repository:** [Sentence Transformers on GitHub](https://github.com/huggingface/sentence-transformers)
-- **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)
-### Full Model Architecture
-```
-SentenceTransformer(
-  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False, 'architecture': 'XLMRobertaModel'})
-  (1): Pooling({'word_embedding_dimension': 768, '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})
-)
-```
-## Usage
-### Direct Usage (Sentence Transformers)
-First install the Sentence Transformers library:
-```bash
-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("sentence_transformers_model_id")
-# Run inference
-sentences = [
-    'Kiedy rozpoczyna się kurs?',
-    'Kurs startuje dla Ciebie w momencie, kiedy się na niego zapiszesz. Chwilę później otrzymasz pierwszą lekcję oraz obiecany do niej BONUS w postaci e-booka „100+ nieoczywistych pomysłów na e-sklep”. Kolejne lekcje będziemy wysyłać co 3 dni.',
-    'Nowe kursy grupowe zaczynają się w połowie lutego. Dokładna data jest przekazana kursantom tydzień przed rozpoczęciem kursu.',
-]
-embeddings = model.encode(sentences)
-print(embeddings.shape)
-# [3, 768]
-# Get the similarity scores for the embeddings
-similarities = model.similarity(embeddings, embeddings)
-print(similarities)
-# tensor([[1.0000, 0.9920, 0.9929],
-#         [0.9920, 1.0000, 0.9964],
-#         [0.9929, 0.9964, 1.0000]])
-```
-<!--
-### Direct Usage (Transformers)
-<details><summary>Click to see the direct usage in Transformers</summary>
-</details>
--->
-<!--
-### Downstream Usage (Sentence Transformers)
-You can finetune this model on your own dataset.
-<details><summary>Click to expand</summary>
-</details>
--->
-<!--
-### Out-of-Scope Use
-*List how the model may foreseeably be misused and address what users ought not to do with the model.*
--->
-<!--
-## Bias, Risks and Limitations
-*What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*
--->
-<!--
-### Recommendations
-*What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*
--->
-## Training Details
-### Training Dataset
-#### Unnamed Dataset
-* Size: 12,753,278 training samples
-* Columns: <code>sentence_0</code>, <code>sentence_1</code>, <code>sentence_2</code>, and <code>label</code>
-* Approximate statistics based on the first 1000 samples:
-  |         | sentence_0                                                                         | sentence_1                                                                         | sentence_2                                                                           | label                                                            |
-  |:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------|:-----------------------------------------------------------------|
-  | type    | string                                                                             | string                                                                             | string                                                                               | float                                                            |
-  | details | <ul><li>min: 6 tokens</li><li>mean: 15.49 tokens</li><li>max: 119 tokens</li></ul> | <ul><li>min: 9 tokens</li><li>mean: 74.63 tokens</li><li>max: 512 tokens</li></ul> | <ul><li>min: 10 tokens</li><li>mean: 102.27 tokens</li><li>max: 512 tokens</li></ul> | <ul><li>min: -0.97</li><li>mean: 0.43</li><li>max: 1.0</li></ul> |
-* Samples:
-  | sentence_0                                                                   | sentence_1                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        | sentence_2                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               | label                            |
-  |:-----------------------------------------------------------------------------|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------|
-  | <code>Как найти актуальное зеркало Betwinner?</code>                         | <code>Букмекер старается обеспечить доступ к сайту, поэтому ссылки на зеркала обновляются ежедневно. Чтобы быть в курсе всех новостей, рекомендуется подписаться на почтовую рассылку и соцсети.</code>                                                                                                                                                                                                                                                                                                                                           | <code>Зеркало BetWinner можно найти в интернете</code>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   | <code>-0.9661979675292969</code> |
-  | <code>Jakie są minimalne zakłady w bakaracie?</code>                         | <code>Minimalny zakład w bakaracie zależy od konkretnej gry, w którą grasz. Mini Baccarat ma zwykle niskie limity zakładów, co czyni go atrakcyjnym dla nowych graczy. Istnieją też wersje gry w bakarata dla high-rollerów, które nakładają wyższy minimalny zakład.</code>                                                                                                                                                                                                                                                                      | <code>Bakarat to jedna z popularniejszych gier hazardowych. To także najprostsza karciana gra kasynowa. Bakarat charakteryzuje się także niską przewagą kasyna nad graczem. Co za tym idzie stawki wygranych w niej nie są wysokie (najlepiej wyceniany jest zakład na remis, ale wtedy przewaga kasyna bardzo wzrasta - 14,44 proc. przy grze 6 taliami kart i 14,36 proc. przy grze 8 taliami kart). Przy zakładzie na gracza kasyno ma przewagę na poziomie 1,24 proc., a przy zakładzie na bankiera - na poziomie 1,06 proc.</code>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  | <code>0.579345703125</code>      |
-  | <code>Come scegliere il massimale assicurazione professionale medici?</code> | <code>Per il momento, non sono ancora entrate in vigore sul massimale minimo per le polizze rc professionale medici. Teniamo conto però di una cosa: se si lavora (e si è lavorato nei dieci anni precedenti) esclusivamente come dipendenti o specializzandi presso l’SSN, dobbiamo sapere che la rivalsa massima dell’SSN sarà plafonata al triplo del reddito annuo lordo del medico.<br>Se invece si lavora in libera professione, non c’è alcun limite. Consigliamo comunque di scegliere massimali non inferiori al milione di euro.</code> | <code>È un modello alternativo al modello standard dell’assicurazione malattia di base. Usufruite delle stesse prestazioni del modello standard dell’assicurazione malattia di base, ma pagate un premio inferiore. In cambio, accettate di consultare in primo luogo il medico di famiglia che avete scelto. Il medico di famiglia, chiamato anche «medico di primo ricorso (MPR)», vi cura e, se necessario, vi indirizza verso uno specialista. Ciò permette di evitare inutili consulti e contribuisce a ridurre i costi sanitari. Se il vostro medico di famiglia o un altro medico vi indirizza verso uno specialista, dovete chiedere al medico di rilasciarvi un attestato, chiamato anche «buono di delega». Alcuni medici ce lo inviano elettronicamente. In caso contrario, potete chiederlo al medico che vi ha raccomandato lo specialista (basta una semplice annotazione firmata, con indicati il tipo di specialista raccomandato e la durata di validità dell’attestato). Potete inviarci tale documento per posta o tramite ...</code> | <code>0.9451904296875</code>     |
-* Loss: [<code>MarginMSELoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#marginmseloss)
-### Training Hyperparameters
-#### Non-Default Hyperparameters
-- `per_device_train_batch_size`: 64
-- `per_device_eval_batch_size`: 64
-- `num_train_epochs`: 1
-- `fp16`: True
-- `multi_dataset_batch_sampler`: round_robin
-#### All Hyperparameters
-<details><summary>Click to expand</summary>
-- `overwrite_output_dir`: False
-- `do_predict`: False
-- `eval_strategy`: no
-- `prediction_loss_only`: True
-- `per_device_train_batch_size`: 64
-- `per_device_eval_batch_size`: 64
-- `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`: 5e-05
-- `weight_decay`: 0.0
-- `adam_beta1`: 0.9
-- `adam_beta2`: 0.999
-- `adam_epsilon`: 1e-08
-- `max_grad_norm`: 1
-- `num_train_epochs`: 1
-- `max_steps`: -1
-- `lr_scheduler_type`: linear
-- `lr_scheduler_kwargs`: {}
-- `warmup_ratio`: 0.0
-- `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`: 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`: 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`: 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}
-- `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`: 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`: 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`: round_robin
-- `router_mapping`: {}
-- `learning_rate_mapping`: {}
-</details>
-### Training Logs
-<details><summary>Click to expand</summary>
-| Epoch  | Step   | Training Loss |
-|:------:|:------:|:-------------:|
-| 0.0025 | 500    | 3.2897        |
-| 0.0050 | 1000   | 0.1515        |
-| 0.0075 | 1500   | 0.1374        |
-| 0.0100 | 2000   | 0.1319        |
-| 0.0125 | 2500   | 0.1322        |
-| 0.0151 | 3000   | 0.1294        |
-| 0.0176 | 3500   | 0.1254        |
-| 0.0201 | 4000   | 0.1234        |
-| 0.0226 | 4500   | 0.1201        |
-| 0.0251 | 5000   | 0.1196        |
-| 0.0276 | 5500   | 0.1215        |
-| 0.0301 | 6000   | 0.1174        |
-| 0.0326 | 6500   | 0.1184        |
-| 0.0351 | 7000   | 0.1176        |
-| 0.0376 | 7500   | 0.1152        |
-| 0.0401 | 8000   | 0.1141        |
-| 0.0427 | 8500   | 0.1137        |
-| 0.0452 | 9000   | 0.1144        |
-| 0.0477 | 9500   | 0.1132        |
-| 0.0502 | 10000  | 0.1123        |
-| 0.0527 | 10500  | 0.1117        |
-| 0.0552 | 11000  | 0.1117        |
-| 0.0577 | 11500  | 0.1102        |
-| 0.0602 | 12000  | 0.109         |
-| 0.0627 | 12500  | 0.1101        |
-| 0.0652 | 13000  | 0.1079        |
-| 0.0677 | 13500  | 0.1106        |
-| 0.0703 | 14000  | 0.1097        |
-| 0.0728 | 14500  | 0.1075        |
-| 0.0753 | 15000  | 0.1046        |
-| 0.0778 | 15500  | 0.1078        |
-| 0.0803 | 16000  | 0.1061        |
-| 0.0828 | 16500  | 0.1057        |
-| 0.0853 | 17000  | 0.1054        |
-| 0.0878 | 17500  | 0.1067        |
-| 0.0903 | 18000  | 0.1048        |
-| 0.0928 | 18500  | 0.1033        |
-| 0.0953 | 19000  | 0.104         |
-| 0.0979 | 19500  | 0.102         |
-| 0.1004 | 20000  | 0.1023        |
-| 0.1029 | 20500  | 0.101         |
-| 0.1054 | 21000  | 0.1035        |
-| 0.1079 | 21500  | 0.102         |
-| 0.1104 | 22000  | 0.1018        |
-| 0.1129 | 22500  | 0.1015        |
-| 0.1154 | 23000  | 0.1003        |
-| 0.1179 | 23500  | 0.1005        |
-| 0.1204 | 24000  | 0.0998        |
-| 0.1229 | 24500  | 0.099         |
-| 0.1255 | 25000  | 0.1001        |
-| 0.1280 | 25500  | 0.0979        |
-| 0.1305 | 26000  | 0.1001        |
-| 0.1330 | 26500  | 0.0995        |
-| 0.1355 | 27000  | 0.0992        |
-| 0.1380 | 27500  | 0.098         |
-| 0.1405 | 28000  | 0.0986        |
-| 0.1430 | 28500  | 0.0987        |
-| 0.1455 | 29000  | 0.0972        |
-| 0.1480 | 29500  | 0.0964        |
-| 0.1505 | 30000  | 0.0967        |
-| 0.1531 | 30500  | 0.0969        |
-| 0.1556 | 31000  | 0.0954        |
-| 0.1581 | 31500  | 0.0972        |
-| 0.1606 | 32000  | 0.0973        |
-| 0.1631 | 32500  | 0.096         |
-| 0.1656 | 33000  | 0.0952        |
-| 0.1681 | 33500  | 0.0974        |
-| 0.1706 | 34000  | 0.0945        |
-| 0.1731 | 34500  | 0.0936        |
-| 0.1756 | 35000  | 0.0945        |
-| 0.1782 | 35500  | 0.0946        |
-| 0.1807 | 36000  | 0.0942        |
-| 0.1832 | 36500  | 0.0955        |
-| 0.1857 | 37000  | 0.0948        |
-| 0.1882 | 37500  | 0.0925        |
-| 0.1907 | 38000  | 0.0929        |
-| 0.1932 | 38500  | 0.0934        |
-| 0.1957 | 39000  | 0.0939        |
-| 0.1982 | 39500  | 0.0933        |
-| 0.2007 | 40000  | 0.0937        |
-| 0.2032 | 40500  | 0.0916        |
-| 0.2058 | 41000  | 0.0932        |
-| 0.2083 | 41500  | 0.0921        |
-| 0.2108 | 42000  | 0.0912        |
-| 0.2133 | 42500  | 0.0906        |
-| 0.2158 | 43000  | 0.0905        |
-| 0.2183 | 43500  | 0.09          |
-| 0.2208 | 44000  | 0.0906        |
-| 0.2233 | 44500  | 0.092         |
-| 0.2258 | 45000  | 0.0906        |
-| 0.2283 | 45500  | 0.0908        |
-| 0.2308 | 46000  | 0.0916        |
-| 0.2334 | 46500  | 0.0907        |
-| 0.2359 | 47000  | 0.0899        |
-| 0.2384 | 47500  | 0.089         |
-| 0.2409 | 48000  | 0.0909        |
-| 0.2434 | 48500  | 0.0889        |
-| 0.2459 | 49000  | 0.0896        |
-| 0.2484 | 49500  | 0.088         |
-| 0.2509 | 50000  | 0.09          |
-| 0.2534 | 50500  | 0.0879        |
-| 0.2559 | 51000  | 0.0885        |
-| 0.2584 | 51500  | 0.0886        |
-| 0.2610 | 52000  | 0.0896        |
-| 0.2635 | 52500  | 0.0886        |
-| 0.2660 | 53000  | 0.0876        |
-| 0.2685 | 53500  | 0.0881        |
-| 0.2710 | 54000  | 0.0886        |
-| 0.2735 | 54500  | 0.0865        |
-| 0.2760 | 55000  | 0.0874        |
-| 0.2785 | 55500  | 0.0878        |
-| 0.2810 | 56000  | 0.0874        |
-| 0.2835 | 56500  | 0.0872        |
-| 0.2860 | 57000  | 0.0866        |
-| 0.2886 | 57500  | 0.0875        |
-| 0.2911 | 58000  | 0.0876        |
-| 0.2936 | 58500  | 0.0872        |
-| 0.2961 | 59000  | 0.0857        |
-| 0.2986 | 59500  | 0.0867        |
-| 0.3011 | 60000  | 0.0862        |
-| 0.3036 | 60500  | 0.0849        |
-| 0.3061 | 61000  | 0.0863        |
-| 0.3086 | 61500  | 0.0849        |
-| 0.3111 | 62000  | 0.0857        |
-| 0.3136 | 62500  | 0.084         |
-| 0.3162 | 63000  | 0.0857        |
-| 0.3187 | 63500  | 0.0853        |
-| 0.3212 | 64000  | 0.0849        |
-| 0.3237 | 64500  | 0.0842        |
-| 0.3262 | 65000  | 0.0851        |
-| 0.3287 | 65500  | 0.085         |
-| 0.3312 | 66000  | 0.0837        |
-| 0.3337 | 66500  | 0.0839        |
-| 0.3362 | 67000  | 0.0836        |
-| 0.3387 | 67500  | 0.0845        |
-| 0.3412 | 68000  | 0.0844        |
-| 0.3438 | 68500  | 0.0844        |
-| 0.3463 | 69000  | 0.0839        |
-| 0.3488 | 69500  | 0.084         |
-| 0.3513 | 70000  | 0.083         |
-| 0.3538 | 70500  | 0.0843        |
-| 0.3563 | 71000  | 0.082         |
-| 0.3588 | 71500  | 0.0834        |
-| 0.3613 | 72000  | 0.0826        |
-| 0.3638 | 72500  | 0.0833        |
-| 0.3663 | 73000  | 0.0843        |
-| 0.3688 | 73500  | 0.0821        |
-| 0.3714 | 74000  | 0.0822        |
-| 0.3739 | 74500  | 0.0823        |
-| 0.3764 | 75000  | 0.0818        |
-| 0.3789 | 75500  | 0.0836        |
-| 0.3814 | 76000  | 0.0813        |
-| 0.3839 | 76500  | 0.0829        |
-| 0.3864 | 77000  | 0.0828        |
-| 0.3889 | 77500  | 0.0799        |
-| 0.3914 | 78000  | 0.0819        |
-| 0.3939 | 78500  | 0.0815        |
-| 0.3964 | 79000  | 0.0812        |
-| 0.3990 | 79500  | 0.0803        |
-| 0.4015 | 80000  | 0.0819        |
-| 0.4040 | 80500  | 0.081         |
-| 0.4065 | 81000  | 0.0798        |
-| 0.4090 | 81500  | 0.0811        |
-| 0.4115 | 82000  | 0.0806        |
-| 0.4140 | 82500  | 0.0812        |
-| 0.4165 | 83000  | 0.0801        |
-| 0.4190 | 83500  | 0.0803        |
-| 0.4215 | 84000  | 0.0812        |
-| 0.4240 | 84500  | 0.0809        |
-| 0.4266 | 85000  | 0.0802        |
-| 0.4291 | 85500  | 0.0801        |
-| 0.4316 | 86000  | 0.08          |
-| 0.4341 | 86500  | 0.079         |
-| 0.4366 | 87000  | 0.0803        |
-| 0.4391 | 87500  | 0.08          |
-| 0.4416 | 88000  | 0.0802        |
-| 0.4441 | 88500  | 0.0799        |
-| 0.4466 | 89000  | 0.0795        |
-| 0.4491 | 89500  | 0.0787        |
-| 0.4516 | 90000  | 0.0784        |
-| 0.4542 | 90500  | 0.0781        |
-| 0.4567 | 91000  | 0.0802        |
-| 0.4592 | 91500  | 0.0781        |
-| 0.4617 | 92000  | 0.0796        |
-| 0.4642 | 92500  | 0.0774        |
-| 0.4667 | 93000  | 0.0794        |
-| 0.4692 | 93500  | 0.0786        |
-| 0.4717 | 94000  | 0.079         |
-| 0.4742 | 94500  | 0.0786        |
-| 0.4767 | 95000  | 0.0778        |
-| 0.4792 | 95500  | 0.0782        |
-| 0.4818 | 96000  | 0.0777        |
-| 0.4843 | 96500  | 0.0773        |
-| 0.4868 | 97000  | 0.0762        |
-| 0.4893 | 97500  | 0.0774        |
-| 0.4918 | 98000  | 0.0796        |
-| 0.4943 | 98500  | 0.0764        |
-| 0.4968 | 99000  | 0.0781        |
-| 0.4993 | 99500  | 0.0778        |
-| 0.5018 | 100000 | 0.0774        |
-| 0.5043 | 100500 | 0.0767        |
-| 0.5069 | 101000 | 0.0769        |
-| 0.5094 | 101500 | 0.0784        |
-| 0.5119 | 102000 | 0.0769        |
-| 0.5144 | 102500 | 0.0773        |
-| 0.5169 | 103000 | 0.0776        |
-| 0.5194 | 103500 | 0.0761        |
-| 0.5219 | 104000 | 0.0768        |
-| 0.5244 | 104500 | 0.0763        |
-| 0.5269 | 105000 | 0.0772        |
-| 0.5294 | 105500 | 0.076         |
-| 0.5319 | 106000 | 0.0776        |
-| 0.5345 | 106500 | 0.0768        |
-| 0.5370 | 107000 | 0.0754        |
-| 0.5395 | 107500 | 0.0759        |
-| 0.5420 | 108000 | 0.0764        |
-| 0.5445 | 108500 | 0.0764        |
-| 0.5470 | 109000 | 0.0766        |
-| 0.5495 | 109500 | 0.0762        |
-| 0.5520 | 110000 | 0.0749        |
-| 0.5545 | 110500 | 0.075         |
-| 0.5570 | 111000 | 0.0754        |
-| 0.5595 | 111500 | 0.0755        |
-| 0.5621 | 112000 | 0.0753        |
-| 0.5646 | 112500 | 0.0747        |
-| 0.5671 | 113000 | 0.0754        |
-| 0.5696 | 113500 | 0.0756        |
-| 0.5721 | 114000 | 0.074         |
-| 0.5746 | 114500 | 0.0759        |
-| 0.5771 | 115000 | 0.0755        |
-| 0.5796 | 115500 | 0.0757        |
-| 0.5821 | 116000 | 0.0744        |
-| 0.5846 | 116500 | 0.0732        |
-| 0.5871 | 117000 | 0.0745        |
-| 0.5897 | 117500 | 0.0748        |
-| 0.5922 | 118000 | 0.0724        |
-| 0.5947 | 118500 | 0.0739        |
-| 0.5972 | 119000 | 0.0749        |
-| 0.5997 | 119500 | 0.0755        |
-| 0.6022 | 120000 | 0.0735        |
-| 0.6047 | 120500 | 0.0742        |
-| 0.6072 | 121000 | 0.0738        |
-| 0.6097 | 121500 | 0.0733        |
-| 0.6122 | 122000 | 0.0728        |
-| 0.6147 | 122500 | 0.0745        |
-| 0.6173 | 123000 | 0.0741        |
-| 0.6198 | 123500 | 0.0726        |
-| 0.6223 | 124000 | 0.0744        |
-| 0.6248 | 124500 | 0.0743        |
-| 0.6273 | 125000 | 0.0732        |
-| 0.6298 | 125500 | 0.0731        |
-| 0.6323 | 126000 | 0.0729        |
-| 0.6348 | 126500 | 0.0737        |
-| 0.6373 | 127000 | 0.0735        |
-| 0.6398 | 127500 | 0.0738        |
-| 0.6423 | 128000 | 0.0731        |
-| 0.6449 | 128500 | 0.0736        |
-| 0.6474 | 129000 | 0.0728        |
-| 0.6499 | 129500 | 0.073         |
-| 0.6524 | 130000 | 0.0733        |
-| 0.6549 | 130500 | 0.073         |
-| 0.6574 | 131000 | 0.073         |
-| 0.6599 | 131500 | 0.0732        |
-| 0.6624 | 132000 | 0.0723        |
-| 0.6649 | 132500 | 0.0732        |
-| 0.6674 | 133000 | 0.0724        |
-| 0.6699 | 133500 | 0.0722        |
-| 0.6725 | 134000 | 0.0724        |
-| 0.6750 | 134500 | 0.0726        |
-| 0.6775 | 135000 | 0.0728        |
-| 0.6800 | 135500 | 0.0717        |
-| 0.6825 | 136000 | 0.0722        |
-| 0.6850 | 136500 | 0.0729        |
-| 0.6875 | 137000 | 0.0715        |
-| 0.6900 | 137500 | 0.072         |
-| 0.6925 | 138000 | 0.072         |
-| 0.6950 | 138500 | 0.0722        |
-| 0.6975 | 139000 | 0.0718        |
-| 0.7001 | 139500 | 0.0728        |
-| 0.7026 | 140000 | 0.0718        |
-| 0.7051 | 140500 | 0.0726        |
-| 0.7076 | 141000 | 0.0707        |
-| 0.7101 | 141500 | 0.072         |
-| 0.7126 | 142000 | 0.0706        |
-| 0.7151 | 142500 | 0.0706        |
-| 0.7176 | 143000 | 0.0708        |
-| 0.7201 | 143500 | 0.0717        |
-| 0.7226 | 144000 | 0.0713        |
-| 0.7251 | 144500 | 0.0723        |
-| 0.7277 | 145000 | 0.0709        |
-| 0.7302 | 145500 | 0.0709        |
-| 0.7327 | 146000 | 0.0706        |
-| 0.7352 | 146500 | 0.0713        |
-| 0.7377 | 147000 | 0.0709        |
-| 0.7402 | 147500 | 0.0703        |
-| 0.7427 | 148000 | 0.0709        |
-| 0.7452 | 148500 | 0.0702        |
-| 0.7477 | 149000 | 0.0705        |
-| 0.7502 | 149500 | 0.0707        |
-| 0.7527 | 150000 | 0.0702        |
-| 0.7553 | 150500 | 0.0696        |
-| 0.7578 | 151000 | 0.0701        |
-| 0.7603 | 151500 | 0.0707        |
-| 0.7628 | 152000 | 0.0703        |
-| 0.7653 | 152500 | 0.0703        |
-| 0.7678 | 153000 | 0.0711        |
-| 0.7703 | 153500 | 0.0706        |
-| 0.7728 | 154000 | 0.0701        |
-| 0.7753 | 154500 | 0.0699        |
-| 0.7778 | 155000 | 0.0704        |
-| 0.7803 | 155500 | 0.07          |
-| 0.7829 | 156000 | 0.0701        |
-| 0.7854 | 156500 | 0.0697        |
-| 0.7879 | 157000 | 0.0698        |
-| 0.7904 | 157500 | 0.0699        |
-| 0.7929 | 158000 | 0.069         |
-| 0.7954 | 158500 | 0.0703        |
-| 0.7979 | 159000 | 0.0696        |
-| 0.8004 | 159500 | 0.0701        |
-| 0.8029 | 160000 | 0.069         |
-| 0.8054 | 160500 | 0.0687        |
-| 0.8079 | 161000 | 0.069         |
-| 0.8105 | 161500 | 0.0692        |
-| 0.8130 | 162000 | 0.069         |
-| 0.8155 | 162500 | 0.0688        |
-| 0.8180 | 163000 | 0.0681        |
-| 0.8205 | 163500 | 0.0688        |
-| 0.8230 | 164000 | 0.0699        |
-| 0.8255 | 164500 | 0.0677        |
-| 0.8280 | 165000 | 0.0687        |
-| 0.8305 | 165500 | 0.0696        |
-| 0.8330 | 166000 | 0.0686        |
-| 0.8355 | 166500 | 0.069         |
-| 0.8381 | 167000 | 0.0692        |
-| 0.8406 | 167500 | 0.0698        |
-| 0.8431 | 168000 | 0.0684        |
-| 0.8456 | 168500 | 0.0681        |
-| 0.8481 | 169000 | 0.0683        |
-| 0.8506 | 169500 | 0.0701        |
-| 0.8531 | 170000 | 0.0697        |
-| 0.8556 | 170500 | 0.0688        |
-| 0.8581 | 171000 | 0.0689        |
-| 0.8606 | 171500 | 0.0689        |
-| 0.8632 | 172000 | 0.0687        |
-| 0.8657 | 172500 | 0.0693        |
-| 0.8682 | 173000 | 0.0678        |
-| 0.8707 | 173500 | 0.0688        |
-| 0.8732 | 174000 | 0.0686        |
-| 0.8757 | 174500 | 0.0695        |
-| 0.8782 | 175000 | 0.0679        |
-| 0.8807 | 175500 | 0.0686        |
-| 0.8832 | 176000 | 0.0683        |
-| 0.8857 | 176500 | 0.068         |
-| 0.8882 | 177000 | 0.0688        |
-| 0.8908 | 177500 | 0.0696        |
-| 0.8933 | 178000 | 0.0682        |
-| 0.8958 | 178500 | 0.0686        |
-| 0.8983 | 179000 | 0.0679        |
-| 0.9008 | 179500 | 0.0687        |
-| 0.9033 | 180000 | 0.0677        |
-| 0.9058 | 180500 | 0.0693        |
-| 0.9083 | 181000 | 0.0685        |
-| 0.9108 | 181500 | 0.0682        |
-| 0.9133 | 182000 | 0.0689        |
-| 0.9158 | 182500 | 0.0682        |
-| 0.9184 | 183000 | 0.0679        |
-| 0.9209 | 183500 | 0.0682        |
-| 0.9234 | 184000 | 0.0678        |
-| 0.9259 | 184500 | 0.0685        |
-| 0.9284 | 185000 | 0.0673        |
-| 0.9309 | 185500 | 0.0676        |
-| 0.9334 | 186000 | 0.068         |
-| 0.9359 | 186500 | 0.0678        |
-| 0.9384 | 187000 | 0.0679        |
-| 0.9409 | 187500 | 0.0674        |
-| 0.9434 | 188000 | 0.068         |
-| 0.9460 | 188500 | 0.0679        |
-| 0.9485 | 189000 | 0.0673        |
-| 0.9510 | 189500 | 0.0663        |
-| 0.9535 | 190000 | 0.068         |
-| 0.9560 | 190500 | 0.0672        |
-| 0.9585 | 191000 | 0.0668        |
-| 0.9610 | 191500 | 0.0665        |
-| 0.9635 | 192000 | 0.0679        |
-| 0.9660 | 192500 | 0.0678        |
-| 0.9685 | 193000 | 0.0667        |
-| 0.9710 | 193500 | 0.068         |
-| 0.9736 | 194000 | 0.0669        |
-| 0.9761 | 194500 | 0.0686        |
-| 0.9786 | 195000 | 0.0682        |
-| 0.9811 | 195500 | 0.0673        |
-| 0.9836 | 196000 | 0.0682        |
-| 0.9861 | 196500 | 0.0675        |
-| 0.9886 | 197000 | 0.0669        |
-| 0.9911 | 197500 | 0.0669        |
-| 0.9936 | 198000 | 0.0686        |
-| 0.9961 | 198500 | 0.068         |
-| 0.9986 | 199000 | 0.0667        |
-</details>
-### Framework Versions
-- Python: 3.10.4
-- Sentence Transformers: 5.2.0
-- Transformers: 4.57.3
-- PyTorch: 2.9.1+cu128
-- Accelerate: 1.12.0
-- Datasets: 2.21.0
-- Tokenizers: 0.22.1
 ## 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",
-}
-```
-#### MarginMSELoss
-```bibtex
-@misc{hofstätter2021improving,
-    title={Improving Efficient Neural Ranking Models with Cross-Architecture Knowledge Distillation},
-    author={Sebastian Hofstätter and Sophia Althammer and Michael Schröder and Mete Sertkan and Allan Hanbury},
-    year={2021},
-    eprint={2010.02666},
-    archivePrefix={arXiv},
-    primaryClass={cs.IR}
-}
-```
-<!--
-## Glossary
-*Clearly define terms in order to be accessible across audiences.*
--->
-<!--
-## Model Card Authors
-*Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*
--->
-<!--
-## Model Card Contact
-*Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*
--->

 ---
+language:
+- ara
+- dan
+- deu
+- eng
+- fas
+- fra
+- hin
+- ind
+- ita
+- jpn
+- kor
+- nld
+- pol
+- por
+- rus
+- spa
+- swe
+- tur
+- vie
+- zho
+multilingual: true
 tags:
+- dense-retrieval
+- hard-negatives
+- knowledge-distillation
+- webfaq
+license: cc-by-4.0
+task_categories:
 - sentence-similarity
+- text-retrieval
 ---
+# WebFAQ 2.0: Multilingual Hard Negatives
+This dataset contains **mined hard negatives** derived from the **WebFAQ 2.0** corpus. It covers roughly **1.3 million** samples across **20 languages**.
+The dataset is designed to support robust training of dense retrieval models, specifically enabling:
+1.  **Contrastive Learning:** Using strict hard negatives to improve discrimination.
+2.  **Knowledge Distillation:** Using the provided cross-encoder scores to train with soft labels (e.g., MarginMSE).
+## Dataset Creation & Mining Process
+To ensure high-quality training signals, we employed a **two-stage mining pipeline** that balances difficulty with correctness.
+### 1. Lexical Retrieval (Recall)
+For every query in WebFAQ, we first retrieved the **top-200 candidate answers** from the monolingual corpus using **BM25**.
+* **Goal:** Identify candidates with high lexical overlap (shared keywords) that are likely to be "hard" for a dense retriever to distinguish.
+### 2. Semantic Reranking (Precision)
+We reranked the top-200 candidates using the state-of-the-art cross-encoder model: **[BAAI/bge-m3](https://huggingface.co/BAAI/bge-m3)**.
+* **Goal:** Assess the true semantic relevance of each candidate.
+### 3. Filtering & Scoring
+We applied a rigorous filtering strategy to curate the final dataset:
+* **False Negative Removal:** Candidates with extremely high cross-encoder scores (semantic matches) were discarded to prevent "poisoning" the training data with valid answers labeled as negatives.
+* **Easy Negative Removal:** Candidates with very low scores were discarded to ensure training efficiency.
+* **Score Retention:** We retained the BGE-M3 relevance scores for every negative, enabling knowledge distillation workflows.
+## Dataset Structure
+Each sample in the dataset contains the following fields:
+| Field | Description |
+| :--- | :--- |
+| `query` | The user question. |
+| `positive` | The ground-truth correct answer. |
+| `negative` | The mined hard negative (non-relevant but similar). |
+| `score` | The **BGE-M3 cross-encoder score** for the `(query, negative)` pair. |
+### Code & Reproduction
+The code used for mining, filtering, and training is available in the official repository:
+* **GitHub Repository:** [Link to your GitHub Repo]
+* **WebFAQ Project:** [OpenWebSearch.EU](https://openwebsearch.eu)
 ## Citation
+If you use this dataset, please cite the WebFAQ 2.0 paper:
 ```bibtex
+@inproceedings{dinzinger2025webfaq,
+  title={WebFAQ: A Multilingual Collection of Natural QA Datasets for Dense Retrieval},
+  author={Dinzinger, Michael and Caspari, Laura and Dastidar, Kanishka Ghosh and Mitrović, Jelena and Granitzer, Michael},
+  booktitle={Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval},
+  year={2025}
+}