DFKI-NLP at SemEval-2024 Task 2: Towards Robust LLMs Using Data Perturbations and MinMax Training
This work addresses robustness in LLMs for clinical trial analysis, but it is incremental as it builds on existing methods with specific data perturbations.
The paper tackled the problem of developing robust models for Natural Language Inference on Clinical Trial Reports by using data perturbations and MinMax training with the Mistral model, achieving competitive performance on the SemEval-2024 NLI4CT task.
The NLI4CT task at SemEval-2024 emphasizes the development of robust models for Natural Language Inference on Clinical Trial Reports (CTRs) using large language models (LLMs). This edition introduces interventions specifically targeting the numerical, vocabulary, and semantic aspects of CTRs. Our proposed system harnesses the capabilities of the state-of-the-art Mistral model, complemented by an auxiliary model, to focus on the intricate input space of the NLI4CT dataset. Through the incorporation of numerical and acronym-based perturbations to the data, we train a robust system capable of handling both semantic-altering and numerical contradiction interventions. Our analysis on the dataset sheds light on the challenging sections of the CTRs for reasoning.