Dataset Artifact CS 371N final project. To enhance model robustness against unintended dataset artifacts in pre-trained models, we employ adversarial attacks on the SQuAD dataset. Adversarial Data in LLMs CSE 6521 final project. Examined the susceptibility of three Transformer-based models against adversarial examples and implemented two mitigation techniques for different real-world conditions. Disease Classifier with Radiology CSE 5524 final project. Explored the impact of anatomical segmentation masks on chest X-ray classification by fine-tuning BiomedCLIP under varied training to enhance model interpretability and diagnostic focus.
