CrossEncoder based on BAAI/bge-reranker-v2-m3

This is a Cross Encoder model finetuned from BAAI/bge-reranker-v2-m3 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: BAAI/bge-reranker-v2-m3
Maximum Sequence Length: 512 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': 'XLMRobertaForSequenceClassification'})
)

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("steerrec/bge-reranker-v2-m3-query-note")
# Get scores for pairs of inputs
pairs = [
    ['infp男和infj女', '谁懂啊啊啊‼️💗INFP×INFJ相处好戳我\n绿老头和小蝴蝶都是高敏易碎人群，消息秒回啊，朋友圈点赞啊，这种细微且偏爱的行为，都能让对方感到安心和减少内耗。\n小蝴蝶的话，有事情就不要藏着掖着，对绿老头也不要有过多的试探行为，有问题就直接问，直接说，只要把事情拖到情绪都上来了，那这问题就更难解决了。因为绿老头只是看起来淡然，但其实他们的内心是很火热躁动的，他们只是不好意思主动表达。绿老头得学会调整，明知道小蝴蝶容易想多内耗，那就多主动表现情绪，不要让小蝴蝶觉得你对他淡淡的，要是能偶尔打个只球，明确的表达一下内心，那就更好了。#心理[话题]# #mbti[话题]# #infp[话题]# #infp日常[话题]# #infp和infj[话题]# #INFJ[话题]# #infj的世界[话题]# #我的日常[话题]# #温暖治愈[话题]# '],
    ['电动车远光灯刺眼反击', '电动车在这些情况下是全责哦！注意了哦！\n骑电动车一定要小心这些情况哦！[暗中观察R]#交通事故理赔[话题]# #交通事故[话题]# #法律求助[话题]# #电动车[话题]# #法律咨询[话题]# #交通安全[话题]# '],
    ['电动车远光灯刺眼反击', '支付宝上这个骑行险有用吗\n#车险[话题]# #电动车骑行险[话题]# #支付宝[话题]# '],
    ['牛奶品牌排名', '注意❗注意❗杯子到货，买就送📢\n杯子已经到货啦📣\n依然还是35.9到手2瓶4斤装的鲜奶，在额外赠送2个mini玻璃杯哦~\n顺丰冷链发货，包邮到家📦\n\xa0#乍甸牛奶[话题]#\xa0\xa0#云南游[话题]#\xa0\xa0#鲜奶[话题]#\xa0\xa0#可爱杯子[话题]#\xa0\xa0#杯子分享[话题]#\xa0\xa0#杯子控必入系列[话题]#\xa0\xa0#我就是个杯子控[话题]#\xa0\xa0#杯子[话题]#\xa0\xa0#鲜奶酸奶怎么挑[话题]#\xa0\xa0#鲜奶推荐[话题]#\xa0\xa0#鲜牛乳[话题]#\xa0\xa0#乍甸牛奶福利种草官[话题]#\xa0\xa0#乍甸牛奶也很好喝[话题]#\xa0\xa0#乍甸鲜奶[话题]#\xa0\t\n'],
    ['mbti人格', 'INFP小蝴蝶🦋请谨慎破防😅\n感觉本infp的确心灵上有些许脆弱，在面对朋友或者家人，别人的一句批评或者不认同，会影响我一天的心情～感觉这个习惯跟刻在骨子里一样，一边安慰自己，却一边焦虑😮\u200d💨很难想象有时候自己却很开朗乐观，其实内心很脆弱，一点就破⊙﹏⊙\n\t\n内容纯属娱乐🌚如有雷同纯属巧合🌚\n请大家对号入座哈哈哈哈😂#MBTI16型人格[话题]# #mbti梗图[话题]# #infp精神世界[话题]# #infp日常[话题]# '],
]
scores = model.predict(pairs)
print(scores)
# [0.3757 0.2134 0.1656 0.179  0.272 ]

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'infp男和infj女',
    [
        '谁懂啊啊啊‼️💗INFP×INFJ相处好戳我\n绿老头和小蝴蝶都是高敏易碎人群，消息秒回啊，朋友圈点赞啊，这种细微且偏爱的行为，都能让对方感到安心和减少内耗。\n小蝴蝶的话，有事情就不要藏着掖着，对绿老头也不要有过多的试探行为，有问题就直接问，直接说，只要把事情拖到情绪都上来了，那这问题就更难解决了。因为绿老头只是看起来淡然，但其实他们的内心是很火热躁动的，他们只是不好意思主动表达。绿老头得学会调整，明知道小蝴蝶容易想多内耗，那就多主动表现情绪，不要让小蝴蝶觉得你对他淡淡的，要是能偶尔打个只球，明确的表达一下内心，那就更好了。#心理[话题]# #mbti[话题]# #infp[话题]# #infp日常[话题]# #infp和infj[话题]# #INFJ[话题]# #infj的世界[话题]# #我的日常[话题]# #温暖治愈[话题]# ',
        '电动车在这些情况下是全责哦！注意了哦！\n骑电动车一定要小心这些情况哦！[暗中观察R]#交通事故理赔[话题]# #交通事故[话题]# #法律求助[话题]# #电动车[话题]# #法律咨询[话题]# #交通安全[话题]# ',
        '支付宝上这个骑行险有用吗\n#车险[话题]# #电动车骑行险[话题]# #支付宝[话题]# ',
        '注意❗注意❗杯子到货，买就送📢\n杯子已经到货啦📣\n依然还是35.9到手2瓶4斤装的鲜奶，在额外赠送2个mini玻璃杯哦~\n顺丰冷链发货，包邮到家📦\n\xa0#乍甸牛奶[话题]#\xa0\xa0#云南游[话题]#\xa0\xa0#鲜奶[话题]#\xa0\xa0#可爱杯子[话题]#\xa0\xa0#杯子分享[话题]#\xa0\xa0#杯子控必入系列[话题]#\xa0\xa0#我就是个杯子控[话题]#\xa0\xa0#杯子[话题]#\xa0\xa0#鲜奶酸奶怎么挑[话题]#\xa0\xa0#鲜奶推荐[话题]#\xa0\xa0#鲜牛乳[话题]#\xa0\xa0#乍甸牛奶福利种草官[话题]#\xa0\xa0#乍甸牛奶也很好喝[话题]#\xa0\xa0#乍甸鲜奶[话题]#\xa0\t\n',
        'INFP小蝴蝶🦋请谨慎破防😅\n感觉本infp的确心灵上有些许脆弱，在面对朋友或者家人，别人的一句批评或者不认同，会影响我一天的心情～感觉这个习惯跟刻在骨子里一样，一边安慰自己，却一边焦虑😮\u200d💨很难想象有时候自己却很开朗乐观，其实内心很脆弱，一点就破⊙﹏⊙\n\t\n内容纯属娱乐🌚如有雷同纯属巧合🌚\n请大家对号入座哈哈哈哈😂#MBTI16型人格[话题]# #mbti梗图[话题]# #infp精神世界[话题]# #infp日常[话题]# ',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Reranking

Datasets: query_note_test and query_note_val
Evaluated with CrossEncoderRerankingEvaluator with these parameters:
```
{
    "at_k": 10
}
```

Metric	query_note_test	query_note_val
map	0.4565	0.4572
mrr@10	0.5284	0.5247
ndcg@10	0.5271	0.5328

Training Details

Training Dataset

Unnamed Dataset

Size: 769,572 training samples
Columns: sentence1, sentence2, and label

Approximate statistics based on the first 100 samples:

	sentence1	sentence2	label
type	string	string	float
modality	text	text
details	min: 5 tokens mean: 7.92 tokens max: 13 tokens	min: 18 tokens mean: 208.42 tokens max: 512 tokens	min: 0.0 mean: 0.0 max: 0.0

Samples:

sentence1	sentence2	label
`infp男和infj女`	谁懂啊啊啊‼️💗INFP×INFJ相处好戳我绿老头和小蝴蝶都是高敏易碎人群，消息秒回啊，朋友圈点赞啊，这种细微且偏爱的行为，都能让对方感到安心和减少内耗。小蝴蝶的话，有事情就不要藏着掖着，对绿老头也不要有过多的试探行为，有问题就直接问，直接说，只要把事情拖到情绪都上来了，那这问题就更难解决了。因为绿老头只是看起来淡然，但其实他们的内心是很火热躁动的，他们只是不好意思主动表达。绿老头得学会调整，明知道小蝴蝶容易想多内耗，那就多主动表现情绪，不要让小蝴蝶觉得你对他淡淡的，要是能偶尔打个只球，明确的表达一下内心，那就更好了。#心理[话题]# #mbti[话题]# #infp[话题]# #infp日常[话题]# #infp和infj[话题]# #INFJ[话题]# #infj的世界[话题]# #我的日常[话题]# #温暖治愈[话题]#	`0.0`
`电动车远光灯刺眼反击`	`电动车在这些情况下是全责哦！注意了哦！骑电动车一定要小心这些情况哦！[暗中观察R]#交通事故理赔[话题]# #交通事故[话题]# #法律求助[话题]# #电动车[话题]# #法律咨询[话题]# #交通安全[话题]#`	`0.0`
`电动车远光灯刺眼反击`	`支付宝上这个骑行险有用吗 #车险[话题]# #电动车骑行险[话题]# #支付宝[话题]#`	`0.0`

Loss: BinaryCrossEntropyLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 32
learning_rate: 0.0001
weight_decay: 0.01
num_train_epochs: 1
lr_scheduler_type: cosine
warmup_ratio: 0.05
seed: 3407
bf16: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
prediction_loss_only: True
per_device_train_batch_size: 32
per_device_eval_batch_size: 8
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: 0.0001
weight_decay: 0.01
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: cosine
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: 3407
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}
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
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: False
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: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss	query_note_test_ndcg@10	query_note_val_ndcg@10
-1	-1	-	0.5206	-
0.7048	16950	0.6932	-	-
0.7069	17000	0.5323	-	-
0.7089	17050	0.5209	-	-
0.7110	17100	0.5281	-	-
0.7131	17150	0.5267	-	-
0.7152	17200	0.5124	-	-
0.7173	17250	0.5199	-	-
0.7193	17300	0.5189	-	-
0.7214	17350	0.5059	-	-
0.7235	17400	0.5326	-	-
0.7256	17450	0.5159	-	-
0.7277	17500	0.5246	-	-
0.7297	17550	0.5128	-	-
0.7318	17600	0.5078	-	-
0.7339	17650	0.4966	-	-
0.7360	17700	0.506	-	-
0.7380	17750	0.499	-	-
0.7401	17800	0.5069	-	-
0.7422	17850	0.5387	-	-
0.7443	17900	0.5124	-	-
0.7464	17950	0.522	-	-
0.7484	18000	0.5103	-	-
0.7505	18050	0.5217	-	-
0.7526	18100	0.4939	-	-
0.7547	18150	0.5151	-	-
0.7568	18200	0.4804	-	-
0.7588	18250	0.4969	-	-
0.7609	18300	0.5277	-	-
0.7630	18350	0.5143	-	-
0.7651	18400	0.5063	-	-
0.7672	18450	0.4899	-	-
0.7692	18500	0.5144	-	-
0.7713	18550	0.528	-	-
0.7734	18600	0.5032	-	-
0.7755	18650	0.4956	-	-
0.7775	18700	0.5144	-	-
0.7796	18750	0.5145	-	-
0.7817	18800	0.4971	-	-
0.7838	18850	0.5188	-	-
0.7859	18900	0.501	-	-
0.7879	18950	0.4892	-	-
0.7900	19000	0.4752	-	-
0.7921	19050	0.4984	-	-
0.7942	19100	0.5001	-	-
0.7963	19150	0.4809	-	-
0.7983	19200	0.5085	-	-
0.8004	19250	0.5122	-	-
0.8025	19300	0.5122	-	-
0.8046	19350	0.4909	-	-
0.8067	19400	0.5341	-	-
0.8087	19450	0.5147	-	-
0.8108	19500	0.5095	-	-
0.8129	19550	0.4945	-	-
0.8150	19600	0.4971	-	-
0.8170	19650	0.4967	-	-
0.8191	19700	0.5108	-	-
0.8212	19750	0.4983	-	-
0.8233	19800	0.5154	-	-
0.8254	19850	0.5214	-	-
0.8274	19900	0.4953	-	-
0.8295	19950	0.5079	-	-
0.8316	20000	0.5252	-	-
0.8337	20050	0.4966	-	-
0.8358	20100	0.492	-	-
0.8378	20150	0.5065	-	-
0.8399	20200	0.4825	-	-
0.8420	20250	0.4879	-	-
0.8441	20300	0.5351	-	-
0.8462	20350	0.4904	-	-
0.8482	20400	0.5141	-	-
0.8503	20450	0.5146	-	-
0.8524	20500	0.508	-	-
0.8545	20550	0.5271	-	-
0.8565	20600	0.5057	-	-
0.8586	20650	0.4757	-	-
0.8607	20700	0.5151	-	-
0.8628	20750	0.486	-	-
0.8649	20800	0.4908	-	-
0.8669	20850	0.5287	-	-
0.8690	20900	0.5223	-	-
0.8711	20950	0.5086	-	-
0.8732	21000	0.5066	-	-
0.8753	21050	0.5042	-	-
0.8773	21100	0.5032	-	-
0.8794	21150	0.5123	-	-
0.8815	21200	0.4825	-	-
0.8836	21250	0.5222	-	-
0.8857	21300	0.5044	-	-
0.8877	21350	0.5034	-	-
0.8898	21400	0.5193	-	-
0.8919	21450	0.4975	-	-
0.8940	21500	0.4754	-	-
0.8960	21550	0.5209	-	-
0.8981	21600	0.5024	-	-
0.9002	21650	0.5206	-	-
0.9023	21700	0.5032	-	-
0.9044	21750	0.5264	-	-
0.9064	21800	0.499	-	-
0.9085	21850	0.4967	-	-
0.9106	21900	0.491	-	-
0.9127	21950	0.5056	-	-
0.9148	22000	0.4996	-	-
0.9168	22050	0.4994	-	-
0.9189	22100	0.5254	-	-
0.9210	22150	0.5034	-	-
0.9231	22200	0.5123	-	-
0.9252	22250	0.4956	-	-
0.9272	22300	0.5194	-	-
0.9293	22350	0.474	-	-
0.9314	22400	0.4842	-	-
0.9335	22450	0.4914	-	-
0.9356	22500	0.4925	-	-
0.9376	22550	0.4938	-	-
0.9397	22600	0.5086	-	-
0.9418	22650	0.457	-	-
0.9439	22700	0.5185	-	-
0.9459	22750	0.5268	-	-
0.9480	22800	0.4872	-	-
0.9501	22850	0.5048	-	-
0.9522	22900	0.5103	-	-
0.9543	22950	0.5236	-	-
0.9563	23000	0.5049	-	-
0.9584	23050	0.5041	-	-
0.9605	23100	0.5066	-	-
0.9626	23150	0.5206	-	-
0.9647	23200	0.4732	-	-
0.9667	23250	0.4881	-	-
0.9688	23300	0.5099	-	-
0.9709	23350	0.5226	-	-
0.9730	23400	0.5322	-	-
0.9751	23450	0.4993	-	-
0.9771	23500	0.4856	-	-
0.9792	23550	0.4727	-	-
0.9813	23600	0.5093	-	-
0.9834	23650	0.5073	-	-
0.9854	23700	0.5153	-	-
0.9875	23750	0.4979	-	-
0.9896	23800	0.4961	-	-
0.9917	23850	0.5093	-	-
0.9938	23900	0.4811	-	-
0.9958	23950	0.5008	-	-
0.9979	24000	0.5151	-	-
1.0	24050	0.5318	-	0.5328
-1	-1	-	0.5271	-

Training Time

Training: 57.2 minutes

Framework Versions

Python: 3.12.3
Sentence Transformers: 5.5.1
Transformers: 4.56.2
PyTorch: 2.11.0+cu128
Accelerate: 1.13.0
Datasets: 4.3.0
Tokenizers: 0.22.2

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",
}

Downloads last month: 33

Safetensors

Model size

0.6B params

Tensor type

F32

Model tree for steerrec/bge-reranker-v2-m3-query-note

Base model

BAAI/bge-reranker-v2-m3

Finetuned

(132)

this model

Paper for steerrec/bge-reranker-v2-m3-query-note

Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

Paper • 1908.10084 • Published Aug 27, 2019 • 19

Evaluation results

Map on query note test
self-reported

0.457
Mrr@10 on query note test
self-reported

0.528
Ndcg@10 on query note test
self-reported

0.527
Map on query note val
self-reported

0.457
Mrr@10 on query note val
self-reported

0.525
Ndcg@10 on query note val
self-reported

0.533