README.md

metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:784827
  - loss:ContrastiveLoss
base_model: BAAI/bge-large-en-v1.5
widget:
  - source_sentence: >-
      Represent this sentence for searching relevant passages: Existing methods
      for anomaly detection on dynamic graphs struggle with capturing complex
      time information in graph structures and generating effective negative
      samples for unsupervised learning. These challenges highlight the need for
      improved methodologies that can address the limitations of current
      approaches in this field.We suggest combining 'a message-passing
      framework' and 
    sentences:
      - a single global model
      - videos
      - sequential polygon generation
  - source_sentence: >-
      Represent this sentence for searching relevant passages: The study
      addresses the need for effective tools that allow both novice and expert
      users to analyze the diversity of news coverage about events. It
      highlights the importance of tailoring the interface to accommodate
      non-expert users while also considering the insights of journalism-savvy
      users, indicating a gap in existing systems that cater to varying levels
      of expertise in news analysis.We suggest combining 'a coordinated
      visualization interface tailored for visualization non-expert users' and 
    sentences:
      - worst-case resource analysis
      - Graph Convolution Networks
      - a text encoder
  - source_sentence: >-
      Represent this sentence for searching relevant passages: The accuracy of
      pixel flows is crucial for achieving high-quality video enhancement, yet
      most prior works focus on estimating dense flows that are generally less
      robust and computationally expensive. This highlights a gap in existing
      methodologies that fail to prioritize accuracy over density, necessitating
      a more efficient approach to flow estimation for video enhancement
      tasks.We suggest combining 'sparse point cloud data' and 
    sentences:
      - a Temporal Eigenvalue Loss
      - diffusion models
      - explicit 3D representations, such as polygonal meshes
  - source_sentence: >-
      Represent this sentence for searching relevant passages: The traditional
      frame of discernment lacks a crucial factor, the sequence of propositions,
      which limits the effectiveness of existing methods to measure uncertainty.
      This gap highlights the need for a more comprehensive approach that can
      better represent the relationships between the elements of the frame of
      discernment.We suggest 'combine the order of propositions and the mass of
      them' inspired by 
    sentences:
      - >-
        the traditional matching-optimization methods where matching is
        introduced to handle large displacements before energy-based
        optimizations
      - encoder-decoder models
      - >-
        In another vein, researchers propose new attention augmentation methods
        to make transformers more accurate, efficient and interpretable
  - source_sentence: >-
      Represent this sentence for searching relevant passages: The study
      addresses the need for effective time series forecasting methods to
      estimate the spread of epidemics, particularly in light of the resurgence
      of COVID-19 cases. It highlights the importance of accurately modeling
      both linear and non-linear features of epidemic data to provide state
      authorities and health officials with reliable short-term forecasts and
      strategies.We suggest combining 'ARIMA' and 
    sentences:
      - Transformers
      - a traditional feature-mixed branch
      - >-
        the human brain is able to efficiently learn effective control
        strategies using limited resources
pipeline_tag: sentence-similarity
library_name: sentence-transformers
license: cc
datasets:
  - noystl/Recombination-Pred
language:
  - en

SentenceTransformer based on BAAI/bge-large-en-v1.5

This is a sentence-transformers model finetuned from BAAI/bge-large-en-v1.5. It maps sentences & paragraphs to a 1024-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: BAAI/bge-large-en-v1.5
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity

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': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

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 SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    "Represent this sentence for searching relevant passages: The study addresses the need for effective time series forecasting methods to estimate the spread of epidemics, particularly in light of the resurgence of COVID-19 cases. It highlights the importance of accurately modeling both linear and non-linear features of epidemic data to provide state authorities and health officials with reliable short-term forecasts and strategies.We suggest combining 'ARIMA' and ",
    'Transformers',
    'the human brain is able to efficiently learn effective control strategies using limited resources',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

Training Details

Training Dataset

Unnamed Dataset

• Size: 784,827 training samples
• Columns: query, answer, and label
• Approximate statistics based on the first 1000 samples:

  	query	answer	label
  type	string	string	int
  details	min: 66 tokens mean: 83.86 tokens max: 99 tokens	min: 3 tokens mean: 8.63 tokens max: 49 tokens	0: ~96.70% 1: ~3.30%

• Samples:

  query	answer	label
  Represent this sentence for searching relevant passages: The study addresses the challenge of action segmentation under weak supervision, where the available ground truth only indicates the presence of actions without providing their temporal ordering or occurrence timing in training videos. This limitation necessitates the development of a method to generate pseudo-ground truth for effective training and improve performance in action segmentation and alignment tasks.We suggest combining 'a Hidden Markov Model' and	a multilayer perceptron	1
  Represent this sentence for searching relevant passages: The study addresses the challenge of action segmentation under weak supervision, where the available ground truth only indicates the presence of actions without providing their temporal ordering or occurrence timing in training videos. This limitation necessitates the development of a method to generate pseudo-ground truth for effective training and improve performance in action segmentation and alignment tasks.We suggest combining 'a Hidden Markov Model' and	global expression information	0
  Represent this sentence for searching relevant passages: The study addresses the challenge of action segmentation under weak supervision, where the available ground truth only indicates the presence of actions without providing their temporal ordering or occurrence timing in training videos. This limitation necessitates the development of a method to generate pseudo-ground truth for effective training and improve performance in action segmentation and alignment tasks.We suggest combining 'a Hidden Markov Model' and	some relevant physical parameters	0

• Loss: ContrastiveLoss with these parameters:

  {
      "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
      "margin": 0.5,
      "size_average": true
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 64
• learning_rate: 2.3351317368662443e-06
• warmup_ratio: 0.11883406097525227
• bf16: True
• prompts: {'query': 'Represent this sentence for searching relevant passages: '}
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 64
• 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: 2.3351317368662443e-06
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 3
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.11883406097525227
• 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: 42
• 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: {'query': 'Represent this sentence for searching relevant passages: '}
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Click to expand

Epoch	Step	Training Loss
0.0082	100	0.0051
0.0163	200	0.0038
0.0245	300	0.0037
0.0326	400	0.0036
0.0408	500	0.0046
0.0489	600	0.0035
0.0571	700	0.0035
0.0652	800	0.0034
0.0734	900	0.0044
0.0815	1000	0.0034
0.0897	1100	0.0035
0.0979	1200	0.0034
0.1060	1300	0.0034
0.1142	1400	0.0045
0.1223	1500	0.0034
0.1305	1600	0.0034
0.1386	1700	0.0033
0.1468	1800	0.0043
0.1549	1900	0.0034
0.1631	2000	0.0033
0.1712	2100	0.0032
0.1794	2200	0.0033
0.1876	2300	0.0044
0.1957	2400	0.0033
0.2039	2500	0.0034
0.2120	2600	0.0033
0.2202	2700	0.0042
0.2283	2800	0.0034
0.2365	2900	0.0033
0.2446	3000	0.0033
0.2528	3100	0.0036
0.2609	3200	0.0039
0.2691	3300	0.0033
0.2773	3400	0.0032
0.2854	3500	0.0034
0.2936	3600	0.0041
0.3017	3700	0.0031
0.3099	3800	0.0032
0.3180	3900	0.0031
0.3262	4000	0.004
0.3343	4100	0.0034
0.3425	4200	0.003
0.3506	4300	0.0032
0.3588	4400	0.0032
0.3670	4500	0.004
0.3751	4600	0.0031
0.3833	4700	0.0033
0.3914	4800	0.0031
0.3996	4900	0.004
0.4077	5000	0.0032
0.4159	5100	0.0031
0.4240	5200	0.0031
0.4322	5300	0.0031
0.4403	5400	0.0039
0.4485	5500	0.003
0.4567	5600	0.003
0.4648	5700	0.0031
0.4730	5800	0.0038
0.4811	5900	0.0031
0.4893	6000	0.0032
0.4974	6100	0.0031
0.5056	6200	0.0033
0.5137	6300	0.0035
0.5219	6400	0.0031
0.5300	6500	0.0031
0.5382	6600	0.0031
0.5464	6700	0.0038
0.5545	6800	0.0031
0.5627	6900	0.003
0.5708	7000	0.0029
0.5790	7100	0.0037
0.5871	7200	0.0033
0.5953	7300	0.0031
0.6034	7400	0.003
0.6116	7500	0.003
0.6198	7600	0.004
0.6279	7700	0.0031
0.6361	7800	0.0031
0.6442	7900	0.0031
0.6524	8000	0.0039
0.6605	8100	0.0029
0.6687	8200	0.003
0.6768	8300	0.0029
0.6850	8400	0.0028
0.6931	8500	0.0036
0.7013	8600	0.0031
0.7095	8700	0.0029
0.7176	8800	0.0028
0.7258	8900	0.0035
0.7339	9000	0.0033
0.7421	9100	0.003
0.7502	9200	0.0028
0.7584	9300	0.0029
0.7665	9400	0.0035
0.7747	9500	0.003
0.7828	9600	0.0028
0.7910	9700	0.0027
0.7992	9800	0.0034
0.8073	9900	0.0032
0.8155	10000	0.003
0.8236	10100	0.0029
0.8318	10200	0.0032
0.8399	10300	0.0032
0.8481	10400	0.003
0.8562	10500	0.0029
0.8644	10600	0.0029
0.8725	10700	0.0033
0.8807	10800	0.003
0.8889	10900	0.0029
0.8970	11000	0.0028
0.9052	11100	0.0035
0.9133	11200	0.003
0.9215	11300	0.0029
0.9296	11400	0.0029
0.9378	11500	0.0029
0.9459	11600	0.0034
0.9541	11700	0.0031
0.9622	11800	0.0028
0.9704	11900	0.003
0.9786	12000	0.0035
0.9867	12100	0.0032
0.9949	12200	0.003
1.0030	12300	0.0033
1.0112	12400	0.0029
1.0193	12500	0.003
1.0275	12600	0.0029
1.0356	12700	0.0036
1.0438	12800	0.003
1.0519	12900	0.0027
1.0601	13000	0.0028
1.0683	13100	0.0028
1.0764	13200	0.0036
1.0846	13300	0.0027
1.0927	13400	0.0028
1.1009	13500	0.0029
1.1090	13600	0.0037
1.1172	13700	0.0029
1.1253	13800	0.0029
1.1335	13900	0.0027
1.1416	14000	0.0033
1.1498	14100	0.003
1.1580	14200	0.0027
1.1661	14300	0.0028
1.1743	14400	0.0026
1.1824	14500	0.0036
1.1906	14600	0.0028
1.1987	14700	0.0027
1.2069	14800	0.0029
1.2150	14900	0.0035
1.2232	15000	0.0027
1.2313	15100	0.0027
1.2395	15200	0.0027
1.2477	15300	0.0028
1.2558	15400	0.0035
1.2640	15500	0.0027
1.2721	15600	0.0027
1.2803	15700	0.0027
1.2884	15800	0.0037
1.2966	15900	0.0027
1.3047	16000	0.0027
1.3129	16100	0.0027
1.3210	16200	0.0028
1.3292	16300	0.0033
1.3374	16400	0.0026
1.3455	16500	0.0025
1.3537	16600	0.0028
1.3618	16700	0.0034
1.3700	16800	0.0027
1.3781	16900	0.0026
1.3863	17000	0.0027
1.3944	17100	0.0033
1.4026	17200	0.0027
1.4107	17300	0.0027
1.4189	17400	0.0026
1.4271	17500	0.0027
1.4352	17600	0.0034
1.4434	17700	0.0027
1.4515	17800	0.0025
1.4597	17900	0.0027
1.4678	18000	0.0031
1.4760	18100	0.0027
1.4841	18200	0.0027
1.4923	18300	0.0027
1.5004	18400	0.0027
1.5086	18500	0.0031
1.5168	18600	0.0025
1.5249	18700	0.0026
1.5331	18800	0.0027
1.5412	18900	0.0035
1.5494	19000	0.0025
1.5575	19100	0.0027
1.5657	19200	0.0026
1.5738	19300	0.0028
1.5820	19400	0.0032
1.5901	19500	0.0025
1.5983	19600	0.0027
1.6065	19700	0.0026
1.6146	19800	0.0034
1.6228	19900	0.0027
1.6309	20000	0.0027
1.6391	20100	0.0028
1.6472	20200	0.0031
1.6554	20300	0.0028
1.6635	20400	0.0025
1.6717	20500	0.0025
1.6798	20600	0.0026
1.6880	20700	0.003
1.6962	20800	0.0029
1.7043	20900	0.0027
1.7125	21000	0.0025
1.7206	21100	0.0029
1.7288	21200	0.0029
1.7369	21300	0.0027
1.7451	21400	0.0026
1.7532	21500	0.0025
1.7614	21600	0.003
1.7696	21700	0.0028
1.7777	21800	0.0024
1.7859	21900	0.0025
1.7940	22000	0.003
1.8022	22100	0.0026
1.8103	22200	0.0027
1.8185	22300	0.0027
1.8266	22400	0.0026
1.8348	22500	0.003
1.8429	22600	0.0029
1.8511	22700	0.0025
1.8593	22800	0.0026
1.8674	22900	0.0031
1.8756	23000	0.0027
1.8837	23100	0.0026
1.8919	23200	0.0025
1.9000	23300	0.0028
1.9082	23400	0.0027
1.9163	23500	0.0027
1.9245	23600	0.0027
1.9326	23700	0.0026
1.9408	23800	0.0031
1.9490	23900	0.0027
1.9571	24000	0.0027
1.9653	24100	0.0026
1.9734	24200	0.0032
1.9816	24300	0.0029
1.9897	24400	0.0026
1.9979	24500	0.0028
2.0060	24600	0.0029
2.0142	24700	0.0026
2.0223	24800	0.0027
2.0305	24900	0.0033
2.0387	25000	0.0026
2.0468	25100	0.0026
2.0550	25200	0.0024
2.0631	25300	0.0026
2.0713	25400	0.0033
2.0794	25500	0.0025
2.0876	25600	0.0026
2.0957	25700	0.0026
2.1039	25800	0.0033
2.1120	25900	0.0025
2.1202	26000	0.0026
2.1284	26100	0.0026
2.1365	26200	0.0025
2.1447	26300	0.0031
2.1528	26400	0.0026
2.1610	26500	0.0025
2.1691	26600	0.0026
2.1773	26700	0.0032
2.1854	26800	0.0026
2.1936	26900	0.0026
2.2017	27000	0.0025
2.2099	27100	0.0032
2.2181	27200	0.0025
2.2262	27300	0.0025
2.2344	27400	0.0024
2.2425	27500	0.0025
2.2507	27600	0.0033
2.2588	27700	0.0024
2.2670	27800	0.0024
2.2751	27900	0.0024
2.2833	28000	0.0033
2.2914	28100	0.0025
2.2996	28200	0.0024
2.3078	28300	0.0026
2.3159	28400	0.0024
2.3241	28500	0.0032
2.3322	28600	0.0025
2.3404	28700	0.0024
2.3485	28800	0.0024
2.3567	28900	0.0032
2.3648	29000	0.0025
2.3730	29100	0.0024
2.3811	29200	0.0024
2.3893	29300	0.0028
2.3975	29400	0.003
2.4056	29500	0.0023
2.4138	29600	0.0025
2.4219	29700	0.0024
2.4301	29800	0.0032
2.4382	29900	0.0025
2.4464	30000	0.0024
2.4545	30100	0.0023
2.4627	30200	0.003
2.4708	30300	0.0024
2.4790	30400	0.0025
2.4872	30500	0.0025
2.4953	30600	0.0025
2.5035	30700	0.0031
2.5116	30800	0.0022
2.5198	30900	0.0024
2.5279	31000	0.0024
2.5361	31100	0.0032
2.5442	31200	0.0024
2.5524	31300	0.0023
2.5605	31400	0.0025
2.5687	31500	0.0024
2.5769	31600	0.0031
2.5850	31700	0.0024
2.5932	31800	0.0024
2.6013	31900	0.0024
2.6095	32000	0.0031
2.6176	32100	0.0025
2.6258	32200	0.0025
2.6339	32300	0.0025
2.6421	32400	0.0027
2.6502	32500	0.0029
2.6584	32600	0.0024
2.6666	32700	0.0023
2.6747	32800	0.0025
2.6829	32900	0.0028
2.6910	33000	0.0026
2.6992	33100	0.0025
2.7073	33200	0.0024
2.7155	33300	0.0025
2.7236	33400	0.0026
2.7318	33500	0.0027
2.7399	33600	0.0025
2.7481	33700	0.0024
2.7563	33800	0.0028
2.7644	33900	0.0025
2.7726	34000	0.0024
2.7807	34100	0.0023
2.7889	34200	0.0027
2.7970	34300	0.0024
2.8052	34400	0.0025
2.8133	34500	0.0024
2.8215	34600	0.0024
2.8297	34700	0.0029
2.8378	34800	0.0027
2.8460	34900	0.0025
2.8541	35000	0.0023
2.8623	35100	0.0029
2.8704	35200	0.0025
2.8786	35300	0.0024
2.8867	35400	0.0024
2.8949	35500	0.0024
2.9030	35600	0.0028
2.9112	35700	0.0026
2.9194	35800	0.0023
2.9275	35900	0.0024
2.9357	36000	0.003
2.9438	36100	0.0025
2.9520	36200	0.0025
2.9601	36300	0.0024
2.9683	36400	0.0028
2.9764	36500	0.0027
2.9846	36600	0.0027
2.9927	36700	0.0025

Framework Versions

• Python: 3.11.2
• Sentence Transformers: 3.3.1
• Transformers: 4.49.0
• PyTorch: 2.5.1+cu124
• Accelerate: 1.0.1
• Datasets: 3.1.0
• Tokenizers: 0.21.0

Citation

BibTeX

@misc{sternlicht2025chimeraknowledgebaseidea,
      title={CHIMERA: A Knowledge Base of Idea Recombination in Scientific Literature}, 
      author={Noy Sternlicht and Tom Hope},
      year={2025},
      eprint={2505.20779},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2505.20779}, 
}

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",
}

ContrastiveLoss

@inproceedings{hadsell2006dimensionality,
    author={Hadsell, R. and Chopra, S. and LeCun, Y.},
    booktitle={2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)},
    title={Dimensionality Reduction by Learning an Invariant Mapping},
    year={2006},
    volume={2},
    number={},
    pages={1735-1742},
    doi={10.1109/CVPR.2006.100}
}

Quick Links

• 🌐 Project
• 📃 Paper
• 🛠️ Code