CrossEncoder based on nomic-ai/modernbert-embed-base

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

CityBehavEx

This checkpoint is the profile-coherence aligner (demographic profile consistency scoring) in CityBehavEx, a scalable, empirically validated LLM-assisted urban mobility simulation platform. It is served alongside CityBehavEx's other CrossEncoder aligners by scripts/serve_aligners.py and referenced directly by repo id in scenario configs (e.g. schedule.alignment_model, activities.alignment_model, profiles.coherence_alignment_model, profiles.ownership_alignment_model, activities.poi_type_alignment_model).

If you use this model, please cite CityBehavEx:

@misc{santos2026citybehavex,
  title     = {CityBehavEx: A Scalable and Empirically Validated LLM-Assisted Urban Simulation Platform},
  author    = {Santos, Gustavo H. and Viana, Aline and Silva, Thiago H.},
  year      = {2026},
  eprint    = {2607.12086},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CL},
  url       = {https://arxiv.org/abs/2607.12086}
}

Model Details

Model Description

Model Type: Cross Encoder
Base model: nomic-ai/modernbert-embed-base
Maximum Sequence Length: 8192 tokens
Number of Output Labels: 1 label
Supported Modality: Text

Model Sources

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

Full Model Architecture

CrossEncoder(
  (0): Transformer({'transformer_task': 'sequence-classification', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'logits'}}, 'module_output_name': 'scores', 'architecture': 'ModernBertForSequenceClassification'})
)

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("cross_encoder_model_id")
# Get scores for pairs of inputs
pairs = [
    ['Chloé is a 17-year-old female working as a manager. They have master or above level education and poor health. They live as: couple with children. \nScore whether this synthetic agent profile is demographically coherent and plausible. Use 0 for impossible or highly inconsistent profiles and 1 for fully coherent profiles.', 'demographically coherent and valid synthetic agent profile'],
    ['Clément is a 26-year-old male working as a technician or associate professional. They have secondary or less level education and good health. They live as: living with parents. \nScore whether this synthetic agent profile is demographically coherent and plausible. Use 0 for impossible or highly inconsistent profiles and 1 for fully coherent profiles.', 'demographically coherent and valid synthetic agent profile'],
    ['Chloé is a 25-year-old female working as a service or sales worker. They have secondary or less level education and very poor health. They live as: couple without children. \nScore whether this synthetic agent profile is demographically coherent and plausible. Use 0 for impossible or highly inconsistent profiles and 1 for fully coherent profiles.', 'demographically coherent and valid synthetic agent profile'],
    ['Théo is a 42-year-old male working as a service or sales worker. They have bachelor level education and good health. They live as: living alone. \nScore whether this synthetic agent profile is demographically coherent and plausible. Use 0 for impossible or highly inconsistent profiles and 1 for fully coherent profiles.', 'demographically coherent and valid synthetic agent profile'],
    ['Camille is a 16-year-old female working as a manager. They have no diploma level education and good health. They live as: couple with children. \nScore whether this synthetic agent profile is demographically coherent and plausible. Use 0 for impossible or highly inconsistent profiles and 1 for fully coherent profiles.', 'demographically coherent and valid synthetic agent profile'],
]
scores = model.predict(pairs)
print(scores)
# [0.0939 0.7422 0.8488 0.9852 0.1007]

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'Chloé is a 17-year-old female working as a manager. They have master or above level education and poor health. They live as: couple with children. \nScore whether this synthetic agent profile is demographically coherent and plausible. Use 0 for impossible or highly inconsistent profiles and 1 for fully coherent profiles.',
    [
        'demographically coherent and valid synthetic agent profile',
        'demographically coherent and valid synthetic agent profile',
        'demographically coherent and valid synthetic agent profile',
        'demographically coherent and valid synthetic agent profile',
        'demographically coherent and valid synthetic agent profile',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Training Details

Training Dataset

Unnamed Dataset

Size: 8,000 training samples
Columns: sentence_0, sentence_1, and label

Approximate statistics based on the first 100 samples:

	sentence_0	sentence_1	label
type	string	string	float
modality	text	text
details	min: 62 tokens mean: 66.98 tokens max: 73 tokens	min: 10 tokens mean: 10.0 tokens max: 10 tokens	min: 0.0 mean: 0.43 max: 1.0

Samples:

sentence_0	sentence_1	label
`Chloé is a 17-year-old female working as a manager. They have master or above level education and poor health. They live as: couple with children. Score whether this synthetic agent profile is demographically coherent and plausible. Use 0 for impossible or highly inconsistent profiles and 1 for fully coherent profiles.`	`demographically coherent and valid synthetic agent profile`	`0.1`
`Clément is a 26-year-old male working as a technician or associate professional. They have secondary or less level education and good health. They live as: living with parents. Score whether this synthetic agent profile is demographically coherent and plausible. Use 0 for impossible or highly inconsistent profiles and 1 for fully coherent profiles.`	`demographically coherent and valid synthetic agent profile`	`0.8`
`Chloé is a 25-year-old female working as a service or sales worker. They have secondary or less level education and very poor health. They live as: couple without children. Score whether this synthetic agent profile is demographically coherent and plausible. Use 0 for impossible or highly inconsistent profiles and 1 for fully coherent profiles.`	`demographically coherent and valid synthetic agent profile`	`0.8`

Loss: BinaryCrossEntropyLoss with these parameters:

{
    "activation_fn": "torch.nn.modules.linear.Identity",
    "pos_weight": null
}

Training Hyperparameters

Non-Default Hyperparameters

num_train_epochs: 10

All Hyperparameters

Click to expand

per_device_train_batch_size: 8
num_train_epochs: 10
max_steps: -1
learning_rate: 5e-05
lr_scheduler_type: linear
lr_scheduler_kwargs: None
warmup_steps: 0
optim: adamw_torch_fused
optim_args: None
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
optim_target_modules: None
gradient_accumulation_steps: 1
average_tokens_across_devices: True
max_grad_norm: 1
label_smoothing_factor: 0.0
bf16: False
fp16: False
bf16_full_eval: False
fp16_full_eval: False
tf32: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
use_liger_kernel: False
liger_kernel_config: None
use_cache: False
neftune_noise_alpha: None
torch_empty_cache_steps: None
auto_find_batch_size: False
log_on_each_node: True
logging_nan_inf_filter: True
include_num_input_tokens_seen: no
log_level: passive
log_level_replica: warning
disable_tqdm: False
project: huggingface
trackio_space_id: None
trackio_bucket_id: None
trackio_static_space_id: None
per_device_eval_batch_size: 8
prediction_loss_only: True
eval_on_start: False
eval_do_concat_batches: True
eval_use_gather_object: False
eval_accumulation_steps: None
include_for_metrics: []
batch_eval_metrics: False
save_only_model: False
save_on_each_node: False
enable_jit_checkpoint: False
push_to_hub: False
hub_private_repo: None
hub_model_id: None
hub_strategy: every_save
hub_always_push: False
hub_revision: None
load_best_model_at_end: False
ignore_data_skip: False
restore_callback_states_from_checkpoint: False
full_determinism: False
seed: 42
data_seed: None
use_cpu: 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
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_pin_memory: True
dataloader_persistent_workers: False
dataloader_prefetch_factor: None
dataloader_multiprocessing_context: None
dataloader_in_order: True
remove_unused_columns: True
label_names: None
train_sampling_strategy: random
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
ddp_static_graph: None
ddp_backend: None
ddp_timeout: 1800
fsdp: None
fsdp_config: None
deepspeed: None
debug: []
skip_memory_metrics: True
do_predict: False
resume_from_checkpoint: None
local_rank: -1
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}
warmup_ratio: None

Training Logs

Epoch	Step	Training Loss
0.5	500	0.4738
1.0	1000	0.4035
1.5	1500	0.3949
2.0	2000	0.3908
2.5	2500	0.3867
3.0	3000	0.3887
3.5	3500	0.3843
4.0	4000	0.3873
4.5	4500	0.3831
5.0	5000	0.3874
5.5	5500	0.3869
6.0	6000	0.3824
6.5	6500	0.3853
7.0	7000	0.3828
7.5	7500	0.3849
8.0	8000	0.3827
8.5	8500	0.3833
9.0	9000	0.3836
9.5	9500	0.3822
10.0	10000	0.3840

Training Time

Training: 9.9 minutes

Framework Versions

Python: 3.12.13
Sentence Transformers: 6.0.1
Transformers: 5.17.0
PyTorch: 2.11.0+cu130
Accelerate: 1.15.0
Datasets: 5.0.1
Tokenizers: 0.23.2

Additional Resources

Training and Finetuning Reranker Models with Sentence Transformers: the end-to-end guide for training or finetuning Cross Encoder (reranker) models.
Multimodal Embedding & Reranker Models with Sentence Transformers: use text, image, audio, and video reranker models through the same API.
Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers: training multimodal Cross Encoders.

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",
}