metadata

language:
  - en
license: apache-2.0
tags:
  - sentence-transformers
  - cross-encoder
  - generated_from_trainer
  - dataset_size:578402
  - loss:BinaryCrossEntropyLoss
base_model: answerdotai/ModernBERT-base
pipeline_tag: text-ranking
library_name: sentence-transformers
metrics:
  - map
  - mrr@10
  - ndcg@10
model-index:
  - name: ModernBERT-base trained on GooAQ
    results:
      - task:
          type: cross-encoder-reranking
          name: Cross Encoder Reranking
        dataset:
          name: gooaq dev
          type: gooaq-dev
        metrics:
          - type: map
            value: 0.7283
            name: Map
          - type: mrr@10
            value: 0.7272
            name: Mrr@10
          - type: ndcg@10
            value: 0.7716
            name: Ndcg@10
      - task:
          type: cross-encoder-reranking
          name: Cross Encoder Reranking
        dataset:
          name: NanoMSMARCO R100
          type: NanoMSMARCO_R100
        metrics:
          - type: map
            value: 0.4489
            name: Map
          - type: mrr@10
            value: 0.4376
            name: Mrr@10
          - type: ndcg@10
            value: 0.5096
            name: Ndcg@10
      - task:
          type: cross-encoder-reranking
          name: Cross Encoder Reranking
        dataset:
          name: NanoNFCorpus R100
          type: NanoNFCorpus_R100
        metrics:
          - type: map
            value: 0.3159
            name: Map
          - type: mrr@10
            value: 0.4737
            name: Mrr@10
          - type: ndcg@10
            value: 0.3176
            name: Ndcg@10
      - task:
          type: cross-encoder-reranking
          name: Cross Encoder Reranking
        dataset:
          name: NanoNQ R100
          type: NanoNQ_R100
        metrics:
          - type: map
            value: 0.4904
            name: Map
          - type: mrr@10
            value: 0.5075
            name: Mrr@10
          - type: ndcg@10
            value: 0.5388
            name: Ndcg@10
      - task:
          type: cross-encoder-nano-beir
          name: Cross Encoder Nano BEIR
        dataset:
          name: NanoBEIR R100 mean
          type: NanoBEIR_R100_mean
        metrics:
          - type: map
            value: 0.4184
            name: Map
          - type: mrr@10
            value: 0.4729
            name: Mrr@10
          - type: ndcg@10
            value: 0.4553
            name: Ndcg@10

ModernBERT-base trained on GooAQ

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.

from sentence_transformers import CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("Oysiyl/reranker-ModernBERT-base-gooaq-bce")
# Get scores for pairs of texts
pairs = [
    ['what do you do with a degree in criminal justice?', "['Police Patrol Officer. Job Description. ... ', 'Criminal Investigators & Special Agents. Job Description. ... ', 'Private Detective or Investigator. ... ', 'First-Line Police Supervisor. ... ', 'Correctional Officer. ... ', 'Probation and Parole Officers. ... ', 'Postsecondary Criminal Justice & Law Enforcement Teachers.']"],
    ['what do you do with a degree in criminal justice?', "['Prison officer. ... ', 'Police officer. ... ', 'Detective. ... ', 'Criminologist. ... ', 'Probation officer. ... ', 'Forensic scientist. ... ', 'Crime Scene investigator. ... ', 'Court reporter.']"],
    ['what do you do with a degree in criminal justice?', "A high school diploma is required to work as a detective. In some cases a bachelor's degree in criminal justice or law enforcement may be needed. Experience in law enforcement is usually required, but the amount varies by employer."],
    ['what do you do with a degree in criminal justice?', "['Court Reporter. ... ', 'Criminal Intelligence Analyst. ... ', 'Forensic Accountant. ... ', 'Police Officer and Police Support Roles. ... ', 'Immigration, Customs and Border Roles. ... ', 'Prison Officer. ... ', 'Probation Officer. ... ', 'Scene of Crime Officer.']"],
    ['what do you do with a degree in criminal justice?', "Crime Scene Investigator Education and Training Although a high school diploma or equivalent is a minimum requirement for some positions, many police departments and law enforcement agencies prefer a minimum of an associate's (two-year) or a bachelor's (four-year) degree in criminal justice or a natural science."],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'what do you do with a degree in criminal justice?',
    [
        "['Police Patrol Officer. Job Description. ... ', 'Criminal Investigators & Special Agents. Job Description. ... ', 'Private Detective or Investigator. ... ', 'First-Line Police Supervisor. ... ', 'Correctional Officer. ... ', 'Probation and Parole Officers. ... ', 'Postsecondary Criminal Justice & Law Enforcement Teachers.']",
        "['Prison officer. ... ', 'Police officer. ... ', 'Detective. ... ', 'Criminologist. ... ', 'Probation officer. ... ', 'Forensic scientist. ... ', 'Crime Scene investigator. ... ', 'Court reporter.']",
        "A high school diploma is required to work as a detective. In some cases a bachelor's degree in criminal justice or law enforcement may be needed. Experience in law enforcement is usually required, but the amount varies by employer.",
        "['Court Reporter. ... ', 'Criminal Intelligence Analyst. ... ', 'Forensic Accountant. ... ', 'Police Officer and Police Support Roles. ... ', 'Immigration, Customs and Border Roles. ... ', 'Prison Officer. ... ', 'Probation Officer. ... ', 'Scene of Crime Officer.']",
        "Crime Scene Investigator Education and Training Although a high school diploma or equivalent is a minimum requirement for some positions, many police departments and law enforcement agencies prefer a minimum of an associate's (two-year) or a bachelor's (four-year) degree in criminal justice or a natural science.",
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Reranking

Dataset: gooaq-dev

Evaluated with CrossEncoderRerankingEvaluator with these parameters:

{
    "at_k": 10,
    "always_rerank_positives": false
}

Metric	Value
map	0.7283 (+0.1972)
mrr@10	0.7272 (+0.2033)
ndcg@10	0.7716 (+0.1804)

Cross Encoder Reranking

Datasets: NanoMSMARCO_R100, NanoNFCorpus_R100 and NanoNQ_R100

Evaluated with CrossEncoderRerankingEvaluator with these parameters:

{
    "at_k": 10,
    "always_rerank_positives": true
}

Metric	NanoMSMARCO_R100	NanoNFCorpus_R100	NanoNQ_R100
map	0.4489 (-0.0407)	0.3159 (+0.0549)	0.4904 (+0.0708)
mrr@10	0.4376 (-0.0399)	0.4737 (-0.0262)	0.5075 (+0.0808)
ndcg@10	0.5096 (-0.0309)	0.3176 (-0.0074)	0.5388 (+0.0382)

Cross Encoder Nano BEIR

Dataset: NanoBEIR_R100_mean

Evaluated with CrossEncoderNanoBEIREvaluator with these parameters:

{
    "dataset_names": [
        "msmarco",
        "nfcorpus",
        "nq"
    ],
    "rerank_k": 100,
    "at_k": 10,
    "always_rerank_positives": true
}

Metric	Value
map	0.4184 (+0.0284)
mrr@10	0.4729 (+0.0049)
ndcg@10	0.4553 (-0.0000)

Training Details

Training Dataset

Unnamed Dataset

Size: 578,402 training samples
Columns: question, answer, and label

Approximate statistics based on the first 1000 samples:

	question	answer	label
type	string	string	int
details	min: 20 characters mean: 43.36 characters max: 95 characters	min: 55 characters mean: 252.47 characters max: 386 characters	0: ~82.80% 1: ~17.20%

Samples:

question	answer	label
`what do you do with a degree in criminal justice?`	`['Police Patrol Officer. Job Description. ... ', 'Criminal Investigators & Special Agents. Job Description. ... ', 'Private Detective or Investigator. ... ', 'First-Line Police Supervisor. ... ', 'Correctional Officer. ... ', 'Probation and Parole Officers. ... ', 'Postsecondary Criminal Justice & Law Enforcement Teachers.']`	`1`
`what do you do with a degree in criminal justice?`	`['Prison officer. ... ', 'Police officer. ... ', 'Detective. ... ', 'Criminologist. ... ', 'Probation officer. ... ', 'Forensic scientist. ... ', 'Crime Scene investigator. ... ', 'Court reporter.']`	`0`
`what do you do with a degree in criminal justice?`	`A high school diploma is required to work as a detective. In some cases a bachelor's degree in criminal justice or law enforcement may be needed. Experience in law enforcement is usually required, but the amount varies by employer.`	`0`

Loss: BinaryCrossEntropyLoss with these parameters:

{
    "activation_fn": "torch.nn.modules.linear.Identity",
    "pos_weight": 5
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
learning_rate: 2e-05
num_train_epochs: 1
warmup_ratio: 0.1
seed: 12
bf16: True
dataloader_num_workers: 4
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: 16
per_device_eval_batch_size: 16
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: 4
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
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	-	0.1374 (-0.4539)	0.0392 (-0.5012)	0.2824 (-0.0427)	0.0371 (-0.4635)	0.1196 (-0.3358)
0.0000	1	1.3234	-	-	-	-	-
0.0277	1000	1.1925	-	-	-	-	-
0.0553	2000	0.9574	-	-	-	-	-
0.0830	3000	0.7563	-	-	-	-	-
0.1106	4000	0.7247	0.7171 (+0.1259)	0.5045 (-0.0359)	0.3093 (-0.0158)	0.5701 (+0.0694)	0.4613 (+0.0059)
0.1383	5000	0.6744	-	-	-	-	-
0.1660	6000	0.6876	-	-	-	-	-
0.1936	7000	0.6446	-	-	-	-	-
0.2213	8000	0.6519	0.7319 (+0.1407)	0.5106 (-0.0298)	0.3346 (+0.0095)	0.6054 (+0.1048)	0.4835 (+0.0282)
0.2490	9000	0.6339	-	-	-	-	-
0.2766	10000	0.6168	-	-	-	-	-
0.3043	11000	0.6005	-	-	-	-	-
0.3319	12000	0.6363	0.7463 (+0.1550)	0.5087 (-0.0317)	0.3075 (-0.0176)	0.5757 (+0.0751)	0.4640 (+0.0086)
0.3596	13000	0.5882	-	-	-	-	-
0.3873	14000	0.5888	-	-	-	-	-
0.4149	15000	0.5824	-	-	-	-	-
0.4426	16000	0.5882	0.7487 (+0.1574)	0.5277 (-0.0127)	0.3215 (-0.0036)	0.5609 (+0.0603)	0.4701 (+0.0147)
0.4702	17000	0.5661	-	-	-	-	-
0.4979	18000	0.5758	-	-	-	-	-
0.5256	19000	0.556	-	-	-	-	-
0.5532	20000	0.5524	0.7556 (+0.1644)	0.5419 (+0.0014)	0.3191 (-0.0059)	0.6132 (+0.1125)	0.4914 (+0.0360)
0.5809	21000	0.5546	-	-	-	-	-
0.6086	22000	0.563	-	-	-	-	-
0.6362	23000	0.5369	-	-	-	-	-
0.6639	24000	0.5492	0.7570 (+0.1658)	0.4779 (-0.0626)	0.2881 (-0.0369)	0.5079 (+0.0073)	0.4246 (-0.0307)
0.6915	25000	0.5443	-	-	-	-	-
0.7192	26000	0.5522	-	-	-	-	-
0.7469	27000	0.5323	-	-	-	-	-
0.7745	28000	0.5167	0.7695 (+0.1782)	0.4941 (-0.0463)	0.3167 (-0.0083)	0.5194 (+0.0188)	0.4434 (-0.0119)
0.8022	29000	0.4998	-	-	-	-	-
0.8299	30000	0.5326	-	-	-	-	-
0.8575	31000	0.5262	-	-	-	-	-
0.8852	32000	0.5136	0.7682 (+0.1769)	0.5004 (-0.0400)	0.3273 (+0.0023)	0.5318 (+0.0311)	0.4531 (-0.0022)
0.9128	33000	0.5145	-	-	-	-	-
0.9405	34000	0.5038	-	-	-	-	-
0.9682	35000	0.5236	-	-	-	-	-
0.9958	36000	0.5347	0.7716 (+0.1804)	0.5096 (-0.0309)	0.3176 (-0.0074)	0.5388 (+0.0382)	0.4553 (-0.0000)
-1	-1	-	0.7716 (+0.1804)	0.5096 (-0.0309)	0.3176 (-0.0074)	0.5388 (+0.0382)	0.4553 (-0.0000)

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.10
Sentence Transformers: 4.1.0
Transformers: 4.51.3
PyTorch: 2.7.0+cu128
Accelerate: 1.7.0
Datasets: 3.6.0
Tokenizers: 0.21.1

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",
}