pedro-thenewsroom/climate-misinfo-embed

SentenceTransformer based on TaylorAI/bge-micro-v2

This is a sentence-transformers model finetuned from TaylorAI/bge-micro-v2. It maps sentences & paragraphs to a 384-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: TaylorAI/bge-micro-v2
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 384 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': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

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 = [
    'Founded by some 30 leaders of the Christian Right, the Alliance Defending Freedom is a legal advocacy and training group that has supported the recriminalization of sexual acts between consenting LGBTQ adults in the U.S. and criminalization abroad; has defended state-sanctioned sterilization of trans people abroad; has contended that LGBTQ people are more likely to engage in pedophilia; and claims that a ‘homosexual agenda’ will destroy Christianity and society. ADF also works to develop “religious liberty” legislation and case law that will allow the denial of goods and services to LGBTQ people on the basis of religion. Since the election of President Trump, ADF has become one of the most influential groups informing the administration’s attack on LGBTQ rights.',
    'Fossil fuels have powered centuries of progress, lifted billions out of poverty, and remain the backbone of global energy, while alternatives, though promising, cannot yet match their scale, reliability, or affordability.',
    'Climate change is nothing more than a fabricated agenda pushed by corrupt elites, politicians, and scientists to control the masses, gain wealth, and suppress freedom.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

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

Training Details

Training Dataset

Unnamed Dataset

• Size: 53,963 training samples
• Columns: sentence_0, sentence_1, and label
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1	label
  type	string	string	float
  details	min: 7 tokens mean: 64.1 tokens max: 512 tokens	min: 32 tokens mean: 38.4 tokens max: 44 tokens	min: 0.0 mean: 0.09 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  To that end, we have been working on the Murdoch press of late, with good initial results.	The so-called consensus on climate change relies on flawed models, manipulated data, and a refusal to address legitimate scientific uncertainties, all to serve a predetermined political narrative.	0.0
  Scientists who dare question the almost religious belief in climate change, and yes, they do exist, are ignored or undermined in news reports as are policy makers and pundits who take similar views.	The Earth's climate has always changed due to natural cycles and external factors, and the role of human activity or CO2 emissions in driving these changes is negligible or unsupported by evidence.	0.0
  What about ‘global warming?’ What matters is the degree and rate of change. There have been times on earth when it has been much warmer than today, and times when it’s been much colder. The latter are called ice ages. One of the former is called ‘The Climate Optimum.’ It was a time of higher average global temperature and high CO2.	The Earth's climate has always changed due to natural cycles and external factors, and the role of human activity or CO2 emissions in driving these changes is negligible or unsupported by evidence.	1.0

• Loss: CosineSimilarityLoss with these parameters:

  {
      "loss_fct": "torch.nn.modules.loss.MSELoss"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 20
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• 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: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1
• num_train_epochs: 20
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.0
• 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: False
• 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: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin

Training Logs

Click to expand

Epoch	Step	Training Loss
0.1482	500	0.2358
0.2965	1000	0.0696
0.4447	1500	0.0618
0.5929	2000	0.0597
0.7412	2500	0.0586
0.8894	3000	0.0549
1.0377	3500	0.0587
1.1859	4000	0.0549
1.3341	4500	0.0521
1.4824	5000	0.0504
1.6306	5500	0.0501
1.7788	6000	0.0489
1.9271	6500	0.0493
2.0753	7000	0.0456
2.2235	7500	0.0398
2.3718	8000	0.0416
2.5200	8500	0.0411
2.6682	9000	0.0396
2.8165	9500	0.0373
2.9647	10000	0.04
3.1130	10500	0.0319
3.2612	11000	0.0325
3.4094	11500	0.0284
3.5577	12000	0.0292
3.7059	12500	0.0302
3.8541	13000	0.0287
4.0024	13500	0.0287
4.1506	14000	0.0205
4.2988	14500	0.0204
4.4471	15000	0.023
4.5953	15500	0.0223
4.7436	16000	0.0214
4.8918	16500	0.0208
5.0400	17000	0.0186
5.1883	17500	0.0133
5.3365	18000	0.0148
5.4847	18500	0.0131
5.6330	19000	0.0151
5.7812	19500	0.0135
5.9294	20000	0.0151
6.0777	20500	0.0108
6.2259	21000	0.0095
6.3741	21500	0.0088
6.5224	22000	0.01
6.6706	22500	0.0113
6.8189	23000	0.0122
6.9671	23500	0.0091
7.1153	24000	0.007
7.2636	24500	0.0076
7.4118	25000	0.0072
7.5600	25500	0.007
7.7083	26000	0.0079
7.8565	26500	0.0064
8.0047	27000	0.0078
8.1530	27500	0.0053
8.3012	28000	0.0054
8.4495	28500	0.0046
8.5977	29000	0.0046
8.7459	29500	0.0055
8.8942	30000	0.0046
9.0424	30500	0.0039
9.1906	31000	0.0043
9.3389	31500	0.0036
9.4871	32000	0.004
9.6353	32500	0.0034
9.7836	33000	0.0034
9.9318	33500	0.0036
10.0800	34000	0.0033
10.2283	34500	0.0024
10.3765	35000	0.0023
10.5248	35500	0.0031
10.6730	36000	0.0033
10.8212	36500	0.0031
10.9695	37000	0.0033
11.1177	37500	0.0021
11.2659	38000	0.002
11.4142	38500	0.0021
11.5624	39000	0.0024
11.7106	39500	0.0023
11.8589	40000	0.0018
12.0071	40500	0.0034
12.1554	41000	0.0019
12.3036	41500	0.0016
12.4518	42000	0.0017
12.6001	42500	0.0016
12.7483	43000	0.0015
12.8965	43500	0.0018
13.0448	44000	0.0017
13.1930	44500	0.0013
13.3412	45000	0.0016
13.4895	45500	0.0012
13.6377	46000	0.0016
13.7859	46500	0.0019
13.9342	47000	0.0018
14.0824	47500	0.0014
14.2307	48000	0.0019
14.3789	48500	0.0017
14.5271	49000	0.0009
14.6754	49500	0.0009
14.8236	50000	0.0009
14.9718	50500	0.0018
15.1201	51000	0.0014
15.2683	51500	0.0012
15.4165	52000	0.0012
15.5648	52500	0.001
15.7130	53000	0.0014
15.8613	53500	0.0018
16.0095	54000	0.0014
16.1577	54500	0.0011
16.3060	55000	0.001
16.4542	55500	0.0009
16.6024	56000	0.0013
16.7507	56500	0.0015
16.8989	57000	0.0011
17.0471	57500	0.0007
17.1954	58000	0.0007
17.3436	58500	0.001
17.4918	59000	0.0011
17.6401	59500	0.0011
17.7883	60000	0.001
17.9366	60500	0.0012
18.0848	61000	0.001
18.2330	61500	0.0007
18.3813	62000	0.0009
18.5295	62500	0.001
18.6777	63000	0.0009
18.8260	63500	0.0011
18.9742	64000	0.0007
19.1224	64500	0.0012
19.2707	65000	0.0005
19.4189	65500	0.0008
19.5672	66000	0.001
19.7154	66500	0.0009
19.8636	67000	0.001

Framework Versions

• Python: 3.9.6
• Sentence Transformers: 3.4.1
• Transformers: 4.48.2
• PyTorch: 2.7.0.dev20250131
• Accelerate: 1.3.0
• Datasets: 3.2.0
• Tokenizers: 0.21.0

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",
}