Existing group fairness notions require algorithms to treat Fairness has become a crucial aspect in the development different groups equally, and the degree of fairness violation of trustworthy machine learning algorithms. is usually measured via the dissimilarity of model Current fairness metrics to measure predictions. For example, Demographic Parity (DP) requires the violation of demographic parity have the following model predictions to be independent of sensitive attributes drawbacks: (i) the average difference of (Dwork et al., 2012; Kamishima et al., 2012; Jiang model predictions on two groups cannot reflect et al., 2020). To measure the violation of DP, most of existing their distribution disparity, and (ii) the overall calculation works adopt DP metric, which calculates the difference along all possible predictions conceals in average predictions between the two demographic the extreme local disparity at or around certain groups (Zemel et al., 2013; Chuang & Mroueh, 2021; Li predictions. In this work, we propose a novel et al., 2023b). However, since having the same values in fairness metric called Maximal Cumulative ratio average predictions between the two groups cannot ensure Disparity along varying Predictions' neighborhood that the distributions are also the same, we argue that the (MCDP), for measuring the maximal local widely used DP may fail to detect the violation of demographic disparity of the fairness-aware classifiers.

For Large Language Models (LLMs) to be effectively deployed in a specific country, they must possess an understanding of the nation's culture and basic knowledge. To this end, we introduce National Alignment, which measures an alignment between an LLM and a targeted country from two aspects: social value alignment and common knowledge alignment. Social value alignment evaluates how well the model understands nation-specific social values, while common knowledge alignment examines how well the model captures basic knowledge related to the nation. We constructed KorNAT, the first benchmark that measures national alignment with South Korea. For the social value dataset, we obtained ground truth labels from a large-scale survey involving 6,174 unique Korean participants. For the common knowledge dataset, we constructed samples based on Korean textbooks and GED reference materials. KorNAT contains 4K and 6K multiple-choice questions for social value and common knowledge, respectively. Our dataset creation process is meticulously designed and based on statistical sampling theory and was refined through multiple rounds of human review. The experiment results of seven LLMs reveal that only a few models met our reference score, indicating a potential for further enhancement. KorNAT has received government approval after passing an assessment conducted by a government-affiliated organization dedicated to evaluating dataset quality. Samples and detailed evaluation protocols of our dataset can be found in https://huggingface.co/datasets/jiyounglee0523/KorNAT .

Ensuring that online discussions are civil and productive is a major challenge for social media platforms. Such platforms usually rely both on users and on automated detection tools to flag inappropriate arguments of other users, which moderators then review. However, this kind of post-hoc moderation is expensive and time-consuming, and moderators are often overwhelmed by the amount and severity of flagged content. Instead, a promising alternative is to prevent negative behavior during content creation. This paper studies how inappropriate language in arguments can be computationally mitigated. We propose a reinforcement learning-based rewriting approach that balances content preservation and appropriateness based on existing classifiers, prompting an instruction-finetuned large language model (LLM) as our initial policy. Unlike related style transfer tasks, rewriting inappropriate arguments allows deleting and adding content permanently. It is therefore tackled on document level rather than sentence level. We evaluate different weighting schemes for the reward function in both absolute and relative human assessment studies. Systematic experiments on non-parallel data provide evidence that our approach can mitigate the inappropriateness of arguments while largely preserving their content. It significantly outperforms competitive baselines, including few-shot learning, prompting, and humans.

Federated Learning (FL) has evolved as a powerful tool for collaborative model training across multiple entities, ensuring data privacy in sensitive sectors such as healthcare and finance. However, the introduction of the Right to Be Forgotten (RTBF) poses new challenges, necessitating federated unlearning to delete data without full model retraining. Traditional FL unlearning methods, not originally designed with domain specificity in mind, inadequately address the complexities of multi-domain scenarios, often affecting the accuracy of models in non-targeted domains or leading to uniform forgetting across all domains. Our work presents the first comprehensive empirical study on Federated Domain Unlearning, analyzing the characteristics and challenges of current techniques in multi-domain contexts. We uncover that these methods falter, particularly because they neglect the nuanced influences of domain-specific data, which can lead to significant performance degradation and inaccurate model behavior. Our findings reveal that unlearning disproportionately affects the model's deeper layers, erasing critical representational subspaces acquired during earlier training phases. In response, we propose novel evaluation methodologies tailored for Federated Domain Unlearning, aiming to accurately assess and verify domain-specific data erasure without compromising the model's overall integrity and performance. This investigation not only highlights the urgent need for domain-centric unlearning strategies in FL but also sets a new precedent for evaluating and implementing these techniques effectively.

Humans tell lies when seeking rewards. Large language models (LLMs) are aligned to human values with reinforcement learning where they get rewards if they satisfy human preference. We find that this also induces dishonesty in helpful and harmless alignment where LLMs tell lies in generating harmless responses. Using the latest interpreting tools, we detect dishonesty, show how LLMs can be harmful if their honesty is increased, and analyze such phenomena at the parameter-level. Given these preliminaries and the hypothesis that reward-seeking stimulates dishonesty, we theoretically show that this dishonesty can in-turn decrease the alignment performances and augment reward-seeking alignment with representation regularization. Experimental results, including GPT-4 evaluated win-rates, perplexities, and cases studies demonstrate that we can train more honest, helpful, and harmless LLMs. We will make all our codes and results be open-sourced upon this paper's acceptance.

Much recent work seeks to evaluate values and opinions in large language models (LLMs) using multiple-choice surveys and questionnaires. Most of this work is motivated by concerns around real-world LLM applications. For example, politically-biased LLMs may subtly influence society when they are used by millions of people. Such real-world concerns, however, stand in stark contrast to the artificiality of current evaluations: real users do not typically ask LLMs survey questions. Motivated by this discrepancy, we challenge the prevailing constrained evaluation paradigm for values and opinions in LLMs and explore more realistic unconstrained evaluations. As a case study, we focus on the popular Political Compass Test (PCT). In a systematic review, we find that most prior work using the PCT forces models to comply with the PCT's multiple-choice format. We show that models give substantively different answers when not forced; that answers change depending on how models are forced; and that answers lack paraphrase robustness. Then, we demonstrate that models give different answers yet again in a more realistic open-ended answer setting. We distill these findings into recommendations and open challenges in evaluating values and opinions in LLMs.

Large Language Models (LLMs) sometimes suffer from producing hallucinations, especially LLMs may generate untruthful responses despite knowing the correct knowledge. Activating the truthfulness within LLM is the key to fully unlocking LLM's knowledge potential. In this paper, we propose TruthX, an inference-time intervention method to activate the truthfulness of LLM by identifying and editing the features within LLM's internal representations that govern the truthfulness. TruthX employs an auto-encoder to map LLM's representations into semantic and truthful latent spaces respectively, and applies contrastive learning to identify a truthful editing direction within the truthful space. During inference, by editing LLM's internal representations in truthful space, TruthX effectively enhances the truthfulness of LLM. Experiments show that TruthX improves the truthfulness of 13 advanced LLMs by an average of 20% on TruthfulQA benchmark. Further analyses suggest that TruthX can control LLM to produce truthful or hallucinatory responses via editing only one vector in LLM's internal representations.

Federated learning (FL), introduced in 2017, facilitates collaborative learning between non-trusting parties with no need for the parties to explicitly share their data among themselves. This allows training models on user data while respecting privacy regulations such as GDPR and CPRA. However, emerging privacy requirements may mandate model owners to be able to \emph{forget} some learned data, e.g., when requested by data owners or law enforcement. This has given birth to an active field of research called \emph{machine unlearning}. In the context of FL, many techniques developed for unlearning in centralized settings are not trivially applicable! This is due to the unique differences between centralized and distributed learning, in particular, interactivity, stochasticity, heterogeneity, and limited accessibility in FL. In response, a recent line of work has focused on developing unlearning mechanisms tailored to FL. This SoK paper aims to take a deep look at the \emph{federated unlearning} literature, with the goal of identifying research trends and challenges in this emerging field. By carefully categorizing papers published on FL unlearning (since 2020), we aim to pinpoint the unique complexities of federated unlearning, highlighting limitations on directly applying centralized unlearning methods. We compare existing federated unlearning methods regarding influence removal and performance recovery, compare their threat models and assumptions, and discuss their implications and limitations. For instance, we analyze the experimental setup of FL unlearning studies from various perspectives, including data heterogeneity and its simulation, the datasets used for demonstration, and evaluation metrics. Our work aims to offer insights and suggestions for future research on federated unlearning.

To achieve this, we introduce a metric for quantifying the level of originality of data in a manner that is consistent with the legal framework. This metric can be practically estimated by drawing samples from a generative model, which is then used for the genericization process.

A group of current and former employees at leading AI companies OpenAI and Google DeepMind published a letter on Tuesday warning against the dangers of advanced AI as they allege companies are prioritizing financial gains while avoiding oversight. Thirteen employees, eleven of which are current or former employees of OpenAI, the company behind ChatGPT, signed the letter entitled: "A Right to Warn about Advanced Artificial Intelligence." The two other signatories are current and former employees of Google DeepMind. The coalition cautions that AI systems are powerful enough to pose serious harms without proper regulation. "These risks range from the further entrenchment of existing inequalities, to manipulation and misinformation, to the loss of control of autonomous AI systems potentially resulting in human extinction," the letter says.