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

Model Details

Model Description

Model Type: Cross Encoder
Base model: answerdotai/ModernBERT-base
Maximum Sequence Length: 8192 tokens
Number of Output Labels: 1 label

Language: en
License: apache-2.0

Model Sources

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

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

python

from sentence_transformers import CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("ayushexel/reranker-ModernBERT-base-gooaq-bce-495000")
# Get scores for pairs of texts
pairs = [
    ['how much coffee is in a mocha frappe from starbucks?', 'The Mocha Frappuccino Light Blended Beverage has slightly more caffeine: Tall - 60 mg. Grande - 95 mg. Venti Iced - 120 mg.'],
    ['how much coffee is in a mocha frappe from starbucks?', "Typically Starbucks Vanilla Bean Frappuccino has no coffee in it at all. I believe it's made with sweetened condensed milk, vanilla bean powder and ice and topped with whipped cream of course! According to the Starbucks menu online their frappuccino has 57 grams of sugar and 59 grams of carbs in one 16 ounce serving!"],
    ['how much coffee is in a mocha frappe from starbucks?', 'It also has 5g of protein and 75mg of caffeine. The Mocha Cookie Crumble Frappucino pours a blend of coffee, milk and ice atop whipped cream and chocolate cookie crumble. The whole thing is topped with a blend of rich mocha sauce and Frappuccino chips.'],
    ['how much coffee is in a mocha frappe from starbucks?', 'There are 460 calories in 1 serving of Starbucks Java Chip Frappuccino Blended Coffee with Whipped Cream (Grande).'],
    ['how much coffee is in a mocha frappe from starbucks?', 'Each 14 fl. oz bottle contains 250 calories, 4.5 grams of fat, 38 grams of sugar, and 150 milligrams of caffeine. So, while certainly not the healthiest option, they are convenient for anyone who loves iced Starbucks lattes. The drink is inspired by the Salted Caramel Mocha served seasonally at Starbucks stores.'],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'how much coffee is in a mocha frappe from starbucks?',
    [
        'The Mocha Frappuccino Light Blended Beverage has slightly more caffeine: Tall - 60 mg. Grande - 95 mg. Venti Iced - 120 mg.',
        "Typically Starbucks Vanilla Bean Frappuccino has no coffee in it at all. I believe it's made with sweetened condensed milk, vanilla bean powder and ice and topped with whipped cream of course! According to the Starbucks menu online their frappuccino has 57 grams of sugar and 59 grams of carbs in one 16 ounce serving!",
        'It also has 5g of protein and 75mg of caffeine. The Mocha Cookie Crumble Frappucino pours a blend of coffee, milk and ice atop whipped cream and chocolate cookie crumble. The whole thing is topped with a blend of rich mocha sauce and Frappuccino chips.',
        'There are 460 calories in 1 serving of Starbucks Java Chip Frappuccino Blended Coffee with Whipped Cream (Grande).',
        'Each 14 fl. oz bottle contains 250 calories, 4.5 grams of fat, 38 grams of sugar, and 150 milligrams of caffeine. So, while certainly not the healthiest option, they are convenient for anyone who loves iced Starbucks lattes. The drink is inspired by the Salted Caramel Mocha served seasonally at Starbucks stores.',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Reranking

Dataset: gooaq-dev
Evaluated with CrossEncoderRerankingEvaluator with these parameters:
json
1{ 2 "at_k": 10, 3 "always_rerank_positives": false 4}

Metric	Value
map	0.5977 (+0.2173)
mrr@10	0.5967 (+0.2262)
ndcg@10	0.6431 (+0.2104)

Cross Encoder Reranking

Datasets: NanoMSMARCO_R100, NanoNFCorpus_R100 and NanoNQ_R100
Evaluated with CrossEncoderRerankingEvaluator with these parameters:
json
1{ 2 "at_k": 10, 3 "always_rerank_positives": true 4}

Metric	NanoMSMARCO_R100	NanoNFCorpus_R100	NanoNQ_R100
map	0.3512 (-0.1384)	0.3739 (+0.1129)	0.3826 (-0.0370)
mrr@10	0.3360 (-0.1415)	0.5320 (+0.0322)	0.3942 (-0.0325)
ndcg@10	0.4135 (-0.1269)	0.4074 (+0.0823)	0.4417 (-0.0590)

Cross Encoder Nano BEIR

Dataset: NanoBEIR_R100_mean
Evaluated with CrossEncoderNanoBEIREvaluator with these parameters:
json
1{ 2 "dataset_names": [ 3 "msmarco", 4 "nfcorpus", 5 "nq" 6 ], 7 "rerank_k": 100, 8 "at_k": 10, 9 "always_rerank_positives": true 10}

Metric	Value
map	0.3692 (-0.0208)
mrr@10	0.4207 (-0.0473)
ndcg@10	0.4208 (-0.0345)

Training Details

Training Dataset

Unnamed Dataset

Size: 2,749,365 training samples
Columns: question, answer, and label
Approximate statistics based on the first 1000 samples:
question answer label
type string string int
details
min: 19 characters
mean: 42.48 characters
max: 86 characters
min: 52 characters
mean: 249.98 characters
max: 382 characters
0: ~81.60%
1: ~18.40%

Samples:

question	answer	label
`how much coffee is in a mocha frappe from starbucks?`	`The Mocha Frappuccino Light Blended Beverage has slightly more caffeine: Tall - 60 mg. Grande - 95 mg. Venti Iced - 120 mg.`	`1`
`how much coffee is in a mocha frappe from starbucks?`	`Typically Starbucks Vanilla Bean Frappuccino has no coffee in it at all. I believe it's made with sweetened condensed milk, vanilla bean powder and ice and topped with whipped cream of course! According to the Starbucks menu online their frappuccino has 57 grams of sugar and 59 grams of carbs in one 16 ounce serving!`	`0`
`how much coffee is in a mocha frappe from starbucks?`	`It also has 5g of protein and 75mg of caffeine. The Mocha Cookie Crumble Frappucino pours a blend of coffee, milk and ice atop whipped cream and chocolate cookie crumble. The whole thing is topped with a blend of rich mocha sauce and Frappuccino chips.`	`0`

Loss: BinaryCrossEntropyLoss with these parameters:
json
1{ 2 "activation_fn": "torch.nn.modules.linear.Identity", 3 "pos_weight": 5 4}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 256
per_device_eval_batch_size: 256
learning_rate: 2e-05
num_train_epochs: 1
warmup_ratio: 0.1
seed: 12
bf16: True
dataloader_num_workers: 12
load_best_model_at_end: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 256
per_device_eval_batch_size: 256
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 2e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 1
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.1
warmup_steps: 0
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
save_safetensors: True
save_on_each_node: False
save_only_model: False
restore_callback_states_from_checkpoint: False
no_cuda: False
use_cpu: False
use_mps_device: False
seed: 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: 12
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}
tp_size: 0
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: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	gooaq-dev_ndcg@10	NanoMSMARCO_R100_ndcg@10	NanoNFCorpus_R100_ndcg@10	NanoNQ_R100_ndcg@10	NanoBEIR_R100_mean_ndcg@10
-1	-1	-	0.1284 (-0.3043)	0.0244 (-0.5160)	0.2350 (-0.0900)	0.0158 (-0.4849)	0.0917 (-0.3636)
0.0001	1	1.2609	-	-	-	-	-
0.0186	200	1.2031	-	-	-	-	-
0.0372	400	1.1489	-	-	-	-	-
0.0559	600	0.8525	-	-	-	-	-
0.0745	800	0.7113	-	-	-	-	-
0.0931	1000	0.676	-	-	-	-	-
0.1117	1200	0.6486	-	-	-	-	-
0.1304	1400	0.6133	-	-	-	-	-
0.1490	1600	0.6005	-	-	-	-	-
0.1676	1800	0.5815	-	-	-	-	-
0.1862	2000	0.5711	-	-	-	-	-
0.2048	2200	0.5572	-	-	-	-	-
0.2235	2400	0.5572	-	-	-	-	-
0.2421	2600	0.5449	-	-	-	-	-
0.2607	2800	0.5342	-	-	-	-	-
0.2793	3000	0.5325	-	-	-	-	-
0.2980	3200	0.5321	-	-	-	-	-
0.3166	3400	0.5182	-	-	-	-	-
0.3352	3600	0.5245	-	-	-	-	-
0.3538	3800	0.5302	-	-	-	-	-
0.3724	4000	0.5095	-	-	-	-	-
0.3911	4200	0.5178	-	-	-	-	-
0.4097	4400	0.4962	-	-	-	-	-
0.4283	4600	0.4988	-	-	-	-	-
0.4469	4800	0.4983	-	-	-	-	-
0.4655	5000	0.4973	-	-	-	-	-
0.4842	5200	0.4876	-	-	-	-	-
0.5028	5400	0.4807	-	-	-	-	-
0.5214	5600	0.4862	-	-	-	-	-
0.5400	5800	0.4784	-	-	-	-	-
0.5587	6000	0.4811	-	-	-	-	-
0.5773	6200	0.4817	-	-	-	-	-
0.5959	6400	0.4706	-	-	-	-	-
0.6145	6600	0.4659	-	-	-	-	-
0.6331	6800	0.4644	-	-	-	-	-
0.6518	7000	0.4764	-	-	-	-	-
0.6704	7200	0.4753	-	-	-	-	-
0.6890	7400	0.4727	-	-	-	-	-
0.7076	7600	0.4693	-	-	-	-	-
0.7263	7800	0.4621	-	-	-	-	-
0.7449	8000	0.4514	-	-	-	-	-
0.7635	8200	0.4561	-	-	-	-	-
0.7821	8400	0.4574	-	-	-	-	-
0.8007	8600	0.4579	-	-	-	-	-
0.8194	8800	0.4478	-	-	-	-	-
0.8380	9000	0.4481	-	-	-	-	-
0.8566	9200	0.4568	-	-	-	-	-
0.8752	9400	0.4455	-	-	-	-	-
0.8939	9600	0.4614	-	-	-	-	-
0.9125	9800	0.4436	-	-	-	-	-
0.9311	10000	0.4482	-	-	-	-	-
0.9497	10200	0.4455	-	-	-	-	-
0.9683	10400	0.4422	-	-	-	-	-
0.9870	10600	0.4506	-	-	-	-	-
-1	-1	-	0.6431 (+0.2104)	0.4135 (-0.1269)	0.4074 (+0.0823)	0.4417 (-0.0590)	0.4208 (-0.0345)

Framework Versions

Python: 3.11.0
Sentence Transformers: 4.0.1
Transformers: 4.50.3
PyTorch: 2.6.0+cu124
Accelerate: 1.5.2
Datasets: 3.5.0
Tokenizers: 0.21.1

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",
}

	question	answer	label
type	string	string	int
details	min: 19 characters mean: 42.48 characters max: 86 characters	min: 52 characters mean: 249.98 characters max: 382 characters	0: ~81.60% 1: ~18.40%