Pre-training large language models (LLMs) on vast text corpora enhances natural language processing capabilities but risks encoding social biases, particularly gender bias. While parameter-modification methods like fine-tuning mitigate bias, they are resource-intensive, unsuitable for closed-source models, and lack adaptability to evolving societal norms. Instruction-based approaches offer flexibility but often compromise general performance on normal tasks. To address these limitations, we propose $\textit{FaIRMaker}$, an automated and model-independent framework that employs an $\textbf{auto-search and refinement}$ paradigm to adaptively generate Fairwords, which act as instructions to reduce gender bias and enhance response quality. $\textit{FaIRMaker}$ enhances the debiasing capacity by enlarging the Fairwords search space while preserving the utility and making it applicable to closed-source models by training a sequence-to-sequence model that adaptively refines Fairwords into effective debiasing instructions when facing gender-related queries and performance-boosting prompts for neutral inputs. Extensive experiments demonstrate that $\textit{FaIRMaker}$ effectively mitigates gender bias while preserving task integrity and ensuring compatibility with both open- and closed-source LLMs.

Integrating electronic health records (EHR) into machine learning-driven clinical research and hospital applications is important, as it harnesses extensive and high-quality patient data to enhance outcome predictions and treatment personalization. Nonetheless, due to privacy and security concerns, the secondary purpose of EHR data is consistently governed and regulated, primarily for research intentions, thereby constraining researchers' access to EHR data. Generating synthetic EHR data with deep learning methods is a viable and promising approach to mitigate privacy concerns, offering not only a supplementary resource for downstream applications but also sidestepping the confidentiality risks associated with real patient data. While prior efforts have concentrated on EHR data synthesis, significant challenges persist in the domain of generating synthetic EHR data: balancing the heterogeneity of real EHR including temporal and non-temporal features, addressing the missing values and irregular measures, and ensuring the privacy of the real data used for model training. Existing works in this domain only focused on solving one or two aforementioned challenges. In this work, we propose IGAMT, an innovative framework to generate privacy-preserved synthetic EHR data that not only maintain high quality with heterogeneous features, missing values, and irregular measures but also balances the privacy-utility trade-off. Extensive experiments prove that IGAMT significantly outperforms baseline architectures in terms of visual resemblance and comparable performance in downstream applications. Ablation case studies also prove the effectiveness of the techniques applied in IGAMT.

Prompt-based learning is a new language model training paradigm that adapts the Pre-trained Language Models (PLMs) to downstream tasks, which revitalizes the performance benchmarks across various natural language processing (NLP) tasks.Instead of using a fixed prompt template to fine-tune the model, some research demonstrates the effectiveness of searching for the prompt via optimization.Such prompt optimization process of prompt-based learning on PLMs also gives insight into generating adversarial prompts to mislead the model, raising concerns about the adversarial vulnerability of this paradigm.Recent studies have shown that universal adversarial triggers (UATs) can be generated to alter not only the predictions of the target PLMs but also the prediction of corresponding Prompt-based Fine-tuning Models (PFMs) under the prompt-based learning paradigm.However, UATs found in previous works are often unreadable tokens or characters and can be easily distinguished from natural texts with adaptive defenses.In this work, we consider the naturalness of the UATs and develop LinkPrompt, an adversarial attack algorithm to generate UATs by a gradientbased beam search algorithm that not only effectively attacks the target PLMs and PFMs but also maintains the naturalness among the trigger tokens.Extensive results demonstrate the effectiveness of LinkPrompt, as well as the transferability of UATs generated by LinkPrompt to open-sourced Large Language Model

Tool-augmented language models (TALMs) overcome the limitations of current language models (LMs), allowing them to leverage external tools to enhance performance. One state-of-the-art example is Toolformer introduced by Meta AI Research, which achieves a broader integration of tool utilization. However, Toolformer faces particular concerns related to the robustness of its predictions in the optimal positioning for API calls. Adversarial perturbations can alter the position of API calls chosen by Toolformer, thus resulting in responses that are not only incorrect but potentially even less accurate than those generated by standard language models. To improve the robustness of Toolformer and fulfill the capability of its toolbox, our focus lies on addressing the potential vulnerabilities that arise from small perturbations in the input or prompt space. To achieve this goal, we plan to study adversarial attacks from both attackers' and defenders' perspectives by first studying the adversarial attack algorithms on the input and prompt space, then proposing the certified robustness to the Toolformer API calls scheduling, which is not only empirically effective but also theory-backed.

Recent studies have shown that neural networks are vulnerable to adversarial examples that are designed by adding small perturbations to clean examples in order to trick the classifier to misclassify. Various approaches based on optimization have been proposed for generating adversarial examples with minimal perturbation. Model training based methods such as Adversarial Transformation Network (ATN) provide a fundamentally new way to directly transform an input into an adversarial example, which promises fast generation of adversarial examples. However, the adversarial examples may have suboptimal quality with significantly large perturbations or low attack success rate at small perturbations. In this article, we propose a distance constrainedAdversarialImitationNetwork (AIN), which enhances ATN and is capable of generating both targeted and untargeted examples with an explicit distance constraint. AIN can not only generate large scale adversarial examples efficiently as achieved in ATN, but also imitate the behavior of state-of-the-art optimization-based methods, hence achieving improved quality. Extensive experiments show that AIN significantly outperforms ATN and other Generative Adversarial Networks (GAN) based methods in the quality of generated adversarial examples, and is much more efficient than optimization based methods while achieving comparable quality.

The robustness and security of natural language processing (NLP) models are significantly important in real-world applications.In the context of text classification tasks, adversarial examples can be designed by substituting words with synonyms under certain semantic and syntactic constraints, such that a well-trained model will give a wrong prediction.Therefore, it is crucial to develop techniques to provide a rigorous and provable robustness guarantee against such attacks.In this paper, we propose WordDP to achieve certified robustness against word substitution attacks in text classification via differential privacy (DP).We establish the connection between DP and adversarial robustness for the first time in the text domain and propose a conceptual exponential mechanism-based algorithm to formally achieve the robustness.We further present a practical simulated exponential mechanism that has efficient inference with certified robustness.We not only provide a rigorous analytic derivation of the certified condition but also experimentally compare the utility of WordDP with existing defense algorithms.The results show that WordDP achieves higher accuracy and more than 30⇥ efficiency improvement over the state-of-theart certified robustness mechanism in typical text classification tasks.