Language models (LMs) have demonstrated remarkable capabilities in NLP, yet adapting them efficiently and robustly to specific tasks remains challenging. As their scale and complexity grow, fine-tuning LMs on labelled data often underutilizes available unlabelled data, leads to overfitting on small task-specific sets, and imposes significant computational costs. These limitations hamper their application to the open-ended landscape of real-world language tasks. This thesis proposes a series of methods to better adapt LMs to downstream applications. First, we explore strategies for extracting task-relevant knowledge from unlabelled data, introducing a novel continued pre-training technique that outperforms state-of-the-art semi-supervised approaches. Next, we present a parameter-efficient fine-tuning method that substantially reduces memory and compute costs while maintaining competitive performance. We also introduce improved supervised fine-tuning methods that enable LMs to better follow instructions, especially when labelled data is scarce, enhancing their performance across a range of NLP tasks, including open-ended generation. Finally, we develop new evaluation methods and benchmarks, such as multi-hop spatial reasoning tasks, to assess LM capabilities and adaptation more comprehensively. Through extensive empirical studies across diverse NLP tasks, our results demonstrate that these approaches substantially improve LM robustness, efficiency, and generalization, making them more adaptable to a broad range of applications. These advances mark a significant step towards more robust and efficient LMs, bringing us closer to the goal of artificial general intelligence.

Large language models (LLMs) have demonstrated remarkable capabilities across a broad spectrum of tasks. They have attracted significant attention and been deployed in numerous downstream applications. Nevertheless, akin to a double-edged sword, LLMs also present potential risks. They could suffer from private data leaks or yield inappropriate, harmful, or misleading content. Additionally, the rapid progress of LLMs raises concerns about the potential emergence of superintelligent systems without adequate safeguards. To effectively capitalize on LLM capacities as well as ensure their safe and beneficial development, it is critical to conduct a rigorous and comprehensive evaluation of LLMs. This survey endeavors to offer a panoramic perspective on the evaluation of LLMs. We categorize the evaluation of LLMs into three major groups: knowledge and capability evaluation, alignment evaluation and safety evaluation. In addition to the comprehensive review on the evaluation methodologies and benchmarks on these three aspects, we collate a compendium of evaluations pertaining to LLMs' performance in specialized domains, and discuss the construction of comprehensive evaluation platforms that cover LLM evaluations on capabilities, alignment, safety, and applicability. We hope that this comprehensive overview will stimulate further research interests in the evaluation of LLMs, with the ultimate goal of making evaluation serve as a cornerstone in guiding the responsible development of LLMs. We envision that this will channel their evolution into a direction that maximizes societal benefit while minimizing potential risks. A curated list of related papers has been publicly available at https://github.com/tjunlp-lab/Awesome-LLMs-Evaluation-Papers.

This dissertation proposes a framework of user-centered security in Natural Language Processing (NLP), and demonstrates how it can improve the accessibility of related research. Accordingly, it focuses on two security domains within NLP with great public interest. First, that of author profiling, which can be employed to compromise online privacy through invasive inferences. Without access and detailed insight into these models' predictions, there is no reasonable heuristic by which Internet users might defend themselves from such inferences. Secondly, that of cyberbullying detection, which by default presupposes a centralized implementation; i.e., content moderation across social platforms. As access to appropriate data is restricted, and the nature of the task rapidly evolves (both through lexical variation, and cultural shifts), the effectiveness of its classifiers is greatly diminished and thereby often misrepresented. Under the proposed framework, we predominantly investigate the use of adversarial attacks on language; i.e., changing a given input (generating adversarial samples) such that a given model does not function as intended. These attacks form a common thread between our user-centered security problems; they are highly relevant for privacy-preserving obfuscation methods against author profiling, and adversarial samples might also prove useful to assess the influence of lexical variation and augmentation on cyberbullying detection.