Unlearning is a critical capability for ensuring privacy, security, and compliance in AI systems, enabling models to forget specific data while retaining overall performance. In this work, we participated in Task 4 of SemEval 2025, which focused on unlearning across three sub-tasks: (1) long-form synthetic creative documents, (2) short-form synthetic biographies containing personally identifiable information, and (3) real documents sampled from the target model’s training dataset. We conducted four experiments, employing Supervised Fine-Tuning (SFT) and Negative Preference Optimization (NPO). Despite achieving good performance on the retain set—data that the model was supposed to remember—our findings demonstrate that these techniques did not perform well on the forget set, where unlearning was required.

Human values are of great concern to social sciences which refer to when people have different beliefs and priorities of what is generally worth striving for and how to do so. This paper presents an approach for human value argument mining using contrastive learning to leverage the isotropy of language models. We fine-tuned DeBERTa-Large in a multi-label classification fashion and achieved an F1 score of 49% for the task, resulting in a rank of 11. Our proposed model provides a valuable tool for analyzing arguments related to human values and highlights the significance of leveraging the isotropy of large language models for identifying human values.