Fine-tuning large language models (LLMs) based on human preferences, commonly achieved through reinforcement learning from human feedback (RLHF), has been effective in improving their performance. However, maintaining LLM safety throughout the fine-tuning process remains a significant challenge, as resolving conflicts between safety and helpfulness can be non-trivial. Typically, the safety alignment of LLM is trained on data with safety-related categories. However, our experiments find that naively increasing the scale of safety training data usually leads the LLMs to an ``overly safe'' state rather than a ``truly safe'' state, boosting the refusal rate through extensive safety-aligned data without genuinely understanding the requirements for safe responses. Such an approach can inadvertently diminish the models' helpfulness. To understand the phenomenon, we first investigate the role of safety data by categorizing them into three different groups, and observe that each group behaves differently as training data scales up. To boost the balance between safety and helpfulness, we propose an Equilibrate RLHF framework including a Fine-grained Data-centric (FDC) approach that achieves better safety alignment even with fewer training data, and an Adaptive Message-wise Alignment (AMA) approach, which selectively highlight the key segments through a gradient masking strategy. Extensive experimental results demonstrate that our approach significantly enhances the safety alignment of LLMs while balancing safety and helpfulness.

In the contemporary digital landscape, media content stands as the foundation for political news analysis, offering invaluable insights sourced from various channels like news articles, social media updates, speeches, and reports. Natural Language Processing (NLP) has revolutionized Political Information Extraction (IE), automating tasks such as Event Extraction (EE) from these diverse media outlets. While traditional NLP methods often necessitate specialized expertise to build rule-based systems or train machine learning models with domain-specific datasets, the emergence of Large Language Models (LLMs) driven by Generative Artificial Intelligence (GenAI) presents a promising alternative. These models offer accessibility, alleviating challenges associated with model construction from scratch and reducing the dependency on extensive datasets during the training phase, thus facilitating rapid implementation. However, challenges persist in handling domain-specific tasks, leading to the development of the Retrieval-Augmented Generation (RAG) framework. RAG enhances LLMs by integrating external data retrieval, enriching their contextual understanding, and expanding their knowledge base beyond pre-existing training data. To illustrate RAG's efficacy, we introduce the Political EE system, specifically tailored to extract political event information from news articles. Understanding these political insights is essential for remaining informed about the latest political advancements, whether on a national or global scale.

Large Language Models (LLMs) are being used for a wide variety of tasks. While they are capable of generating human-like responses, they can also produce undesirable output including potentially harmful information, racist or sexist language, and hallucinations. Alignment methods are designed to reduce such undesirable output, via techniques such as fine-tuning, prompt engineering, and representation engineering. However, existing methods face several challenges: some require costly fine-tuning for every alignment task; some do not adequately remove undesirable concepts, failing alignment; some remove benign concepts, lowering the linguistic capabilities of LLMs. To address these issues, we propose Parsimonious Concept Engineering (PaCE), a novel activation engineering framework for alignment. First, to sufficiently model the concepts, we construct a large-scale concept dictionary in the activation space, in which each atom corresponds to a semantic concept. Then, given any alignment task, we instruct a concept partitioner to efficiently annotate the concepts as benign or undesirable. Finally, at inference time, we decompose the LLM activations along the concept dictionary via sparse coding, to accurately represent the activation as a linear combination of the benign and undesirable components. By removing the latter ones from the activation, we reorient the behavior of LLMs towards alignment goals. We conduct experiments on tasks such as response detoxification, faithfulness enhancement, and sentiment revising, and show that PaCE achieves state-of-the-art alignment performance while maintaining linguistic capabilities.