ModernBERT Environment Claims Classifier

This model is a fine-tuned version of answerdotai/ModernBERT-base trained on the QuotaClimat FrugalAIChallenge dataset.

Training Details

The model was trained using the following configuration:

training_args = TrainingArguments(
   output_dir="ModernBERT-envclaims-v0",
   per_device_train_batch_size=32,
   per_device_eval_batch_size=16,
   learning_rate=2e-5,
   num_train_epochs=3,
   bf16=True,
   optim="adamw_torch_fused",
   
   # Logging & Evaluation
   logging_strategy="steps", 
   logging_steps=100,
   eval_strategy="epoch",
   save_strategy="epoch", 
   save_total_limit=2,
   load_best_model_at_end=True,
   metric_for_best_model="f1",
   
   # Training optimization
   weight_decay=0.01,
   lr_scheduler_type="cosine",
   warmup_ratio=0.1,
   
   # Hub parameters
   push_to_hub=True,
   hub_strategy="every_save"
)

Model Performance

The model achieved an F1 score of 0.714 on the evaluation set.

Usage

You can use this model directly with the Hugging Face Transformers library:

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    modelcamillebrl/ModernBERT-envclaims-v0"
)
text = "Your claim here"
class_predicted = classifier(text)

The model classifies texts into the following categories:

Label 0: not_relevant
Label 1: not_happening
Label 2: not_human
Label 3: not_bad
Label 4: solutions_harmful_unnecessary
Label 5: science_unreliable
Label 6: proponents_biased
Label 7: fossil_fuels_needed