This is a sentence-transformers model finetuned from answerdotai/ModernBERT-base on the natural-questions dataset. 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.

This is a test model to experiment with the proposed DebiasedMultipleNegativesRankingLoss from Pull Request #3148 in the Sentence Transformers repository, using commit 370bf473e60b57f7d01a6e084b5acaabdac38a2c.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: answerdotai/ModernBERT-base
Maximum Sequence Length: 8192 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- natural-questions
Language: en

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': 8192, 'do_lower_case': False}) with Transformer model: ModernBertModel 
  (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:

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("tomaarsen/ModernBERT-base-nq-debiased-mnrl")
# Run inference
sentences = [
    'what is meaning of am and pm in time',
    '12-hour clock The 12-hour clock is a time convention in which the 24 hours of the day are divided into two periods:[1] a.m. (from the Latin, ante meridiem, meaning before midday) and p.m. (post meridiem, meaning past midday).[2] Each period consists of 12 hours numbered: 12 (acting as zero),[3] 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, and 11. The 24 hour/day cycle starts at 12 midnight (often indicated as 12 a.m.), runs through 12 noon (often indicated as 12 p.m.), and continues to the midnight at the end of the day. The 12-hour clock was developed over time from the mid-second millennium BC to the 16th century AD.',
    "America's Got Talent America's Got Talent (often abbreviated as AGT) is a televised American talent show competition, broadcast on the NBC television network. It is part of the global Got Talent franchise created by Simon Cowell, and is produced by Fremantle North America and SYCOtv, with distribution done by Fremantle. Since its premiere in June 2006, each season is run during the network's summer schedule, with the show having featured various hosts - it is currently hosted by Tyra Banks, since 2017.[2] It is the first global edition of the franchise, after plans for a British edition in 2005 were suspended, following a dispute between Paul O'Grady, the planned host, and the British broadcaster ITV; production of this edition later resumed in 2007.[3]",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

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

Evaluation

Metrics

Information Retrieval

Datasets: NanoMSMARCO and NanoHotpotQA
Evaluated with InformationRetrievalEvaluator

Metric	NanoMSMARCO	NanoHotpotQA
cosine_accuracy@1	0.14	0.14
cosine_accuracy@3	0.24	0.28
cosine_accuracy@5	0.3	0.3
cosine_accuracy@10	0.4	0.36
cosine_precision@1	0.14	0.14
cosine_precision@3	0.08	0.0933
cosine_precision@5	0.06	0.064
cosine_precision@10	0.04	0.038
cosine_recall@1	0.14	0.07
cosine_recall@3	0.24	0.14
cosine_recall@5	0.3	0.16
cosine_recall@10	0.4	0.19
cosine_ndcg@10	0.2508	0.1572
cosine_mrr@10	0.2056	0.2122
cosine_map@100	0.2194	0.1232

Nano BEIR

Dataset: NanoBEIR_mean
Evaluated with NanoBEIREvaluator

Metric	Value
cosine_accuracy@1	0.14
cosine_accuracy@3	0.26
cosine_accuracy@5	0.3
cosine_accuracy@10	0.38
cosine_precision@1	0.14
cosine_precision@3	0.0867
cosine_precision@5	0.062
cosine_precision@10	0.039
cosine_recall@1	0.105
cosine_recall@3	0.19
cosine_recall@5	0.23
cosine_recall@10	0.295
cosine_ndcg@10	0.204
cosine_mrr@10	0.2089
cosine_map@100	0.1713

Training Details

Training Dataset

natural-questions

Dataset: natural-questions at f9e894e
Size: 100,231 training samples
Columns: query and answer
Approximate statistics based on the first 1000 samples:
query answer
type string string
details
min: 10 tokens
mean: 12.46 tokens
max: 25 tokens
min: 16 tokens
mean: 139.02 tokens
max: 537 tokens

Samples:

query	answer
`who is required to report according to the hmda`	`Home Mortgage Disclosure Act US financial institutions must report HMDA data to their regulator if they meet certain criteria, such as having assets above a specific threshold. The criteria is different for depository and non-depository institutions and are available on the FFIEC website.[4] In 2012, there were 7,400 institutions that reported a total of 18.7 million HMDA records.[5]`
`what is the definition of endoplasmic reticulum in biology`	Endoplasmic reticulum The endoplasmic reticulum (ER) is a type of organelle in eukaryotic cells that forms an interconnected network of flattened, membrane-enclosed sacs or tube-like structures known as cisternae. The membranes of the ER are continuous with the outer nuclear membrane. The endoplasmic reticulum occurs in most types of eukaryotic cells, but is absent from red blood cells and spermatozoa. There are two types of endoplasmic reticulum: rough and smooth. The outer (cytosolic) face of the rough endoplasmic reticulum is studded with ribosomes that are the sites of protein synthesis. The rough endoplasmic reticulum is especially prominent in cells such as hepatocytes. The smooth endoplasmic reticulum lacks ribosomes and functions in lipid manufacture and metabolism, the production of steroid hormones, and detoxification.[1] The smooth ER is especially abundant in mammalian liver and gonad cells. The lacy membranes of the endoplasmic reticulum were first seen in 1945 using elect...
`what does the ski mean in polish names`	`Polish name Since the High Middle Ages, Polish-sounding surnames ending with the masculine -ski suffix, including -cki and -dzki, and the corresponding feminine suffix -ska/-cka/-dzka were associated with the nobility (Polish szlachta), which alone, in the early years, had such suffix distinctions.[1] They are widely popular today.`

Loss: DebiasedMultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 1.0, 3 "similarity_fct": "cos_sim" 4}

Evaluation Dataset

natural-questions

Dataset: natural-questions at f9e894e
Size: 100,231 evaluation samples
Columns: query and answer
Approximate statistics based on the first 1000 samples:
query answer
type string string
details
min: 10 tokens
mean: 12.46 tokens
max: 22 tokens
min: 12 tokens
mean: 138.0 tokens
max: 649 tokens

Samples:

query	answer
`difference between russian blue and british blue cat`	Russian Blue The coat is known as a "double coat", with the undercoat being soft, downy and equal in length to the guard hairs, which are an even blue with silver tips. However, the tail may have a few very dull, almost unnoticeable stripes. The coat is described as thick, plush and soft to the touch. The feeling is softer than the softest silk. The silver tips give the coat a shimmering appearance. Its eyes are almost always a dark and vivid green. Any white patches of fur or yellow eyes in adulthood are seen as flaws in show cats.[3] Russian Blues should not be confused with British Blues (which are not a distinct breed, but rather a British Shorthair with a blue coat as the British Shorthair breed itself comes in a wide variety of colors and patterns), nor the Chartreux or Korat which are two other naturally occurring breeds of blue cats, although they have similar traits.
`who played the little girl on mrs doubtfire`	`Mara Wilson Mara Elizabeth Wilson[2] (born July 24, 1987) is an American writer and former child actress. She is known for playing Natalie Hillard in Mrs. Doubtfire (1993), Susan Walker in Miracle on 34th Street (1994), Matilda Wormwood in Matilda (1996) and Lily Stone in Thomas and the Magic Railroad (2000). Since retiring from film acting, Wilson has focused on writing.`
`what year did the movie the sound of music come out`	The Sound of Music (film) The film was released on March 2, 1965 in the United States, initially as a limited roadshow theatrical release. Although critical response to the film was widely mixed, the film was a major commercial success, becoming the number one box office movie after four weeks, and the highest-grossing film of 1965. By November 1966, The Sound of Music had become the highest-grossing film of all-time—surpassing Gone with the Wind—and held that distinction for five years. The film was just as popular throughout the world, breaking previous box-office records in twenty-nine countries. Following an initial theatrical release that lasted four and a half years, and two successful re-releases, the film sold 283 million admissions worldwide and earned a total worldwide gross of $286,000,000.

Loss: DebiasedMultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 1.0, 3 "similarity_fct": "cos_sim" 4}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 128
per_device_eval_batch_size: 128
learning_rate: 8e-05
num_train_epochs: 1
warmup_ratio: 0.05
seed: 12
bf16: 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: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 8e-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: 1
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: 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: 12
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: 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}
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: 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
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
dispatch_batches: None
split_batches: 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	Validation Loss	NanoMSMARCO_cosine_ndcg@10	NanoHotpotQA_cosine_ndcg@10	NanoBEIR_mean_cosine_ndcg@10
0	0	-	-	0.0785	0.1489	0.1137
0.0129	10	4.8033	-	-	-	-
0.0258	20	4.5295	-	-	-	-
0.0387	30	4.2124	-	-	-	-
0.0515	40	4.0863	-	-	-	-
0.0644	50	4.0048	3.9563	0.1444	0.1660	0.1552
0.0773	60	3.9686	-	-	-	-
0.0902	70	3.9192	-	-	-	-
0.1031	80	3.9276	-	-	-	-
0.1160	90	3.9104	-	-	-	-
0.1289	100	3.8971	3.8877	0.2041	0.1293	0.1667
0.1418	110	3.8987	-	-	-	-
0.1546	120	3.8861	-	-	-	-
0.1675	130	3.8987	-	-	-	-
0.1804	140	3.8811	-	-	-	-
0.1933	150	3.8697	3.8478	0.1918	0.1084	0.1501
0.2062	160	3.8621	-	-	-	-
0.2191	170	3.8628	-	-	-	-
0.2320	180	3.8733	-	-	-	-
0.2448	190	3.8551	-	-	-	-
0.2577	200	3.862	3.8324	0.1940	0.0977	0.1458
0.2706	210	3.8545	-	-	-	-
0.2835	220	3.8495	-	-	-	-
0.2964	230	3.8459	-	-	-	-
0.3093	240	3.8438	-	-	-	-
0.3222	250	3.8425	3.8238	0.1933	0.1498	0.1716
0.3351	260	3.843	-	-	-	-
0.3479	270	3.8486	-	-	-	-
0.3608	280	3.8409	-	-	-	-
0.3737	290	3.8345	-	-	-	-
0.3866	300	3.8446	3.8154	0.1937	0.1532	0.1735
0.3995	310	3.8281	-	-	-	-
0.4124	320	3.8316	-	-	-	-
0.4253	330	3.8325	-	-	-	-
0.4381	340	3.8298	-	-	-	-
0.4510	350	3.8379	3.8104	0.1690	0.1559	0.1624
0.4639	360	3.821	-	-	-	-
0.4768	370	3.8297	-	-	-	-
0.4897	380	3.8206	-	-	-	-
0.5026	390	3.8222	-	-	-	-
0.5155	400	3.8243	3.8031	0.2141	0.1544	0.1843
0.5284	410	3.8328	-	-	-	-
0.5412	420	3.8211	-	-	-	-
0.5541	430	3.82	-	-	-	-
0.5670	440	3.8167	-	-	-	-
0.5799	450	3.8062	3.7988	0.2281	0.1392	0.1837
0.5928	460	3.8166	-	-	-	-
0.6057	470	3.8164	-	-	-	-
0.6186	480	3.8207	-	-	-	-
0.6314	490	3.815	-	-	-	-
0.6443	500	3.813	3.7943	0.2381	0.1260	0.1821
0.6572	510	3.8144	-	-	-	-
0.6701	520	3.8172	-	-	-	-
0.6830	530	3.8175	-	-	-	-
0.6959	540	3.8126	-	-	-	-
0.7088	550	3.8077	3.7913	0.2501	0.1395	0.1948
0.7216	560	3.8022	-	-	-	-
0.7345	570	3.8131	-	-	-	-
0.7474	580	3.8067	-	-	-	-
0.7603	590	3.8175	-	-	-	-
0.7732	600	3.8084	3.7870	0.2751	0.1480	0.2116
0.7861	610	3.8029	-	-	-	-
0.7990	620	3.8125	-	-	-	-
0.8119	630	3.817	-	-	-	-
0.8247	640	3.8038	-	-	-	-
0.8376	650	3.8054	3.7877	0.2274	0.1449	0.1861
0.8505	660	3.8041	-	-	-	-
0.8634	670	3.8012	-	-	-	-
0.8763	680	3.8117	-	-	-	-
0.8892	690	3.8098	-	-	-	-
0.9021	700	3.8008	3.7848	0.2466	0.1551	0.2008
0.9149	710	3.8038	-	-	-	-
0.9278	720	3.7949	-	-	-	-
0.9407	730	3.8044	-	-	-	-
0.9536	740	3.7982	-	-	-	-
0.9665	750	3.804	3.7832	0.2585	0.1587	0.2086
0.9794	760	3.8038	-	-	-	-
0.9923	770	3.8046	-	-	-	-
1.0	776	-	-	0.2508	0.1572	0.2040

Framework Versions

Python: 3.11.10
Sentence Transformers: 3.4.0.dev0
Transformers: 4.48.0.dev0
PyTorch: 2.6.0.dev20241112+cu121
Accelerate: 1.2.0
Datasets: 3.2.0
Tokenizers: 0.21.0

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",
}

DebiasedMultipleNegativesRankingLoss

bibtex

@inproceedings{chuang2020debiased,
    title={Debiased Contrastive Learning},
    author={Ching-Yao Chuang and Joshua Robinson and Lin Yen-Chen and Antonio Torralba and Stefanie Jegelka},
    booktitle={Advances in Neural Information Processing Systems},
    year={2020},
    url={https://arxiv.org/pdf/2007.00224}
}

	query	answer
type	string	string
details	min: 10 tokens mean: 12.46 tokens max: 25 tokens	min: 16 tokens mean: 139.02 tokens max: 537 tokens

tomaarsen/test-ModernBERT-base-nq-debiased-mnrl

SentenceTransformer based on answerdotai/ModernBERT-base

Model Details

Model Description

Model Sources

Full Model Architecture

Usage

Direct Usage (Sentence Transformers)

Evaluation

Metrics

Information Retrieval

Nano BEIR

Training Details

Training Dataset

natural-questions

Evaluation Dataset

natural-questions

Training Hyperparameters

Non-Default Hyperparameters

All Hyperparameters

Training Logs

Framework Versions

Citation

BibTeX

Sentence Transformers

DebiasedMultipleNegativesRankingLoss

tomaarsen/
test-ModernBERT-base-nq-debiased-mnrl