RoBERTa-es-m-large-ICE-MR24-DD is a Spanish language model for mental disorder detection, trained using the Incremental Context Expansion (ICE) methodology. The model was developed for the MentalRisk 2024 Disorder Detection (MR24-DD) shared task and builds upon the RoBERTa-es-mental-large foundation model.

The ICE methodology restructures the training data at the context level, enabling the model to learn from progressively expanding user message histories rather than complete user timelines. This approach allows the model to effectively learn from early and partial evidence while remaining competitive in full-context (user-level) evaluation settings.

The model is based on a RoBERTa-large architecture and supports input sequences of up to 512 tokens, making it suitable for disorder detection scenarios where contextual evidence is incrementally available. It has been fine-tuned to classify users according to the presence of mental health disorders using Spanish user-generated content.

Developed by: ELiRF group, VRAIN (Valencian Research Institute for Artificial Intelligence), Universitat Politècnica de València
Shared by: ELiRF
Model type: Transformer-based sequence classification model (RoBERTa)
Language: Spanish
Base model: RoBERTa-es-mental-large
License: Same as base model

Uses

This model is intended for research purposes in mental health disorder detection.

Direct Use

The model can be used directly for mental disorder detection from Spanish user-generated content, generating predictions from incrementally available user message histories.

Downstream Use

Mental disorder detection (user-level classification)
Early and incremental mental health screening
Comparative studies on disorder detection approaches
Research on temporally-aware and incremental NLP methods

Out-of-Scope Use

Automated intervention systems without human supervision
Use on languages other than Spanish
High-stakes decision-making affecting individuals’ health or safety

ICE Methodology

Incremental Context Expansion (ICE) is a training methodology designed for early and incremental mental health detection tasks. Rather than training on complete user histories, ICE generates multiple incremental contexts per user, each corresponding to a partial message history.

This methodology allows the model to:

Learn from incomplete and early evidence
Improve robustness under early and incremental evaluation metrics
Better align training conditions with real-world deployment scenarios

ICE modifies the dataset construction process while keeping the standard fine-tuning pipeline unchanged.

Bias, Risks, and Limitations

Training data originates from social media platforms and may contain demographic, cultural, and linguistic biases.
Automatically translated texts may introduce translation artifacts or subtle semantic shifts.
Disorder detection is inherently subjective and context-dependent.
The model does not provide explanations or clinical interpretations of its predictions.

How to Get Started with the Model

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("ELiRF/RoBERTa-es-m-large-ICE-MR24-DD")
model = AutoModelForSequenceClassification.from_pretrained(
    "ELiRF/RoBERTa-es-m-large-ICE-MR24-DD"
)

inputs = tokenizer(
    "Ejemplo de historial de mensajes relacionado con salud mental.",
    return_tensors="pt",
    truncation=True,
    max_length=512
)

outputs = model(**inputs)

Training Details

Training Data

The model was fine-tuned on the MentalRisk 2024 Disorder Detection (MR24-DD) dataset. Training data was restructured using the ICE methodology, generating incremental user contexts from original user timelines.

Training Procedure

Base model: RoBERTa-es-mental-large
Fine-tuning strategy: ICE-based context-level training
Objective: Sequence classification
Training regime: fp16 mixed precision

Evaluation

Results

When evaluated on the MentalRisk 2024 Disorder Detection task, RoBERTa-es-m-large-ICE-MR24-DD shows competitive performance and improves upon the state of the art under both full-context (user-level) and early / incremental detection evaluation settings.

Environmental Impact

Hardware type: NVIDIA A40 GPUs
Training time: several hours (fine-tuning)

Technical Specifications

Model Architecture and Objective

Architecture: RoBERTa (large)
Objective: Sequence classification
Maximum sequence length: 512 tokens
Model size: approximately 355M parameters

Citation

This model is part of an ongoing research project. The associated paper is currently under review and will be added to this model card once the publication process is completed.

Model Card Authors

ELiRF research group (VRAIN, Universitat Politècnica de València)