Large Language Models (LLMs) are powerful in-context learners, achieving strong performance with just a few high-quality demonstrations. However, fairness concerns arise in many in-context classification tasks, especially when predictions involve sensitive attributes. To address this, we propose JUDGE—a simple yet effective framework for selecting fair and representative demonstrations that improve group fairness in In-Context Learning. JUDGE constructs the demonstration set iteratively using a greedy approach, guided by a small, carefully selected jury set. Our method remains robust across varying LLM architectures and datasets, ensuring consistent fairness improvements. We evaluate JUDGE on four datasets using four LLMs, comparing it against seven baselines. Results show that JUDGE consistently improves fairness metrics without compromising accuracy.

The widespread popularity of Large Language Models (LLMs), partly due to their emerging in-context learning ability, has highlighted the importance of ethical and safety considerations for deployment. Motivated by corresponding data protection guidelines, we investigate machine unlearning for LLMs. In contrast to the growing literature on fine-tuning methods to achieve unlearning, we focus on a comparatively lightweight alternative called soft prompting to realize unlearning in LLMs. With losses designed to enforce forgetting as well as utility preservation, our framework Soft Prompting for Unlearning (SPUL) learns prompt tokens that are prepended to a query to induce unlearning of specific training examples at inference time without updating LLM parameters. We conduct a rigorous evaluation of the proposed method, and results indicate that SPUL can significantly improve the trade-off between utility and forgetting for text classification and question-answering. We further validate our method with LLMs of varying parameter sizes to highlight its flexibility and provide detailed insights into the choice of hyperparameters and the influence of the size of unlearning data.