Differentially private (DP) mechanisms are difficult to interpret and calibrate because existing methods for mapping standard privacy parameters to concrete privacy risks -- re-identification, attribute inference, and data reconstruction -- are both overly pessimistic and inconsistent. In this work, we use the hypothesis-testing interpretation of DP ($f$-DP), and determine that bounds on attack success can take the same unified form across re-identification, attribute inference, and data reconstruction risks. Our unified bounds are (1) consistent across a multitude of attack settings, and (2) tunable, enabling practitioners to evaluate risk with respect to arbitrary (including worst-case) levels of baseline risk. Empirically, our results are tighter than prior methods using $\varepsilon$-DP, Rényi DP, and concentrated DP. As a result, calibrating noise using our bounds can reduce the required noise by 20% at the same risk level, which yields, e.g., more than 15pp accuracy increase in a text classification task. Overall, this unifying perspective provides a principled framework for interpreting and calibrating the degree of protection in DP against specific levels of re-identification, attribute inference, or data reconstruction risk.

Online comment sections, such as those on news sites or social media, have the potential to foster informal public deliberation, However, this potential is often undermined by the frequency of toxic or low-quality exchanges that occur in these settings. To combat this, platforms increasingly leverage algorithmic ranking to facilitate higher-quality discussions, e.g., by using civility classifiers or forms of prosocial ranking. Yet, these interventions may also inadvertently reduce the visibility of legitimate viewpoints, undermining another key aspect of deliberation: representation of diverse views. We seek to remedy this problem by introducing guarantees of representation into these methods. In particular, we adopt the notion of justified representation (JR) from the social choice literature and incorporate a JR constraint into the comment ranking setting. We find that enforcing JR leads to greater inclusion of diverse viewpoints while still being compatible with optimizing for user engagement or other measures of conversational quality.

Legal judgment prediction (LJP) offers a compelling method to aid legal practitioners and researchers. However, the research question remains relatively underexplored: Should multiple defendants and charges be treated separately in LJP? To address this, we introduce a new dataset, namely multi-person multi-charge prediction (MPMCP), and seek the answer by evaluating the performance of several prevailing legal large language models (LLMs) on four practical legal judgment scenarios: (S1) single defendant with a single charge, (S2) single defendant with multiple charges, (S3) multiple defendants with a single charge, and (S4) multiple defendants with multiple charges. We evaluate the dataset across two LJP tasks, i.e., charge prediction and penalty term prediction. We have conducted extensive experiments and found that the scenario involving multiple defendants and multiple charges (S4) poses the greatest challenges, followed by S2, S3, and S1. The impact varies significantly depending on the model. For example, in S4 compared to S1, InternLM2 achieves approximately 4.5% lower F1-score and 2.8% higher LogD, while Lawformer demonstrates around 19.7% lower F1-score and 19.0% higher LogD.

The rapid expansion of memes on social media has highlighted the urgent need for effective approaches to detect harmful content. However, traditional data-driven approaches struggle to detect new memes due to their evolving nature and the lack of up-to-date annotated data. To address this issue, we propose MIND, a multi-agent framework for zero-shot harmful meme detection that does not rely on annotated data. MIND implements three key strategies: 1) We retrieve similar memes from an unannotated reference set to provide contextual information. 2) We propose a bi-directional insight derivation mechanism to extract a comprehensive understanding of similar memes. 3) We then employ a multi-agent debate mechanism to ensure robust decision-making through reasoned arbitration. Extensive experiments on three meme datasets demonstrate that our proposed framework not only outperforms existing zero-shot approaches but also shows strong generalization across different model architectures and parameter scales, providing a scalable solution for harmful meme detection. The code is available at https://github.com/destroy-lonely/MIND.

Public product launches in Artificial Intelligence can serve as focusing events for collective attention, surfacing how societies react to technological change. Social media provide a window into the sensemaking around these events, surfacing hopes and fears and showing who chooses to engage in the discourse and when. We demonstrate that public sensemaking about AI is shaped by economic interests and cultural values of those involved. We analyze 3.8 million tweets posted by 1.6 million users across 117 countries in response to the public launch of ChatGPT in 2022. Our analysis shows how economic self-interest, proxied by occupational skill types in writing, programming, and mathematics, and national cultural orientations, as measured by Hofstede's individualism, uncertainty avoidance, and power distance dimensions, shape who speaks, when they speak, and their stance towards ChatGPT. Roles requiring more technical skills, such as programming and mathematics, tend to engage earlier and express more positive stances, whereas writing-centric occupations join later with greater skepticism. At the cultural level, individualism predicts both earlier engagement and a more negative stance, and uncertainty avoidance reduces the prevalence of positive stances but does not delay when users first engage with ChatGPT. Aggregate sentiment trends mask the dynamics observed in our study. The shift toward a more critical stance towards ChatGPT over time stems primarily from the entry of more skeptical voices rather than a change of heart among early adopters. Our findings underscore the importance of both the occupational background and cultural context in understanding public reactions to AI.

This project introduces a new measure of elite polarization via actor and subject detection using artificial intelligence. I identify when politicians mention one another in parliamentary speeches, note who is speaking and who is being addressed, and assess the emotional temperature behind these evaluations. This maps how elites evaluate their various out-parties, allowing us to create an index of mutual out-party hostility, that is, elite polarization. While I analyzed polarization data over the past four decades for the UK, and two decades for Hungary and Italy, my approach lays the groundwork for a twenty-year, EU-wide time-series dataset on elite polarization. I obtain the results that can be aggregated by party and quarter. The resulting index demonstrates a good face validity: it reacts to events such as electoral campaigns, country- and party-level crises, and to parties losing and assuming power.

Understanding how policy language evolves over time is critical for assessing global responses to complex challenges such as climate change. Temporal analysis helps stakeholders, including policymakers and researchers, to evaluate past priorities, identify emerging themes, design governance strategies, and develop mitigation measures. Traditional approaches, such as manual thematic coding, are time-consuming and limited in capturing the complex, interconnected nature of global policy discourse. With the increasing relevance of unsupervised machine learning, these limitations can be addressed, particularly under high-volume, complex, and high-dimensional data conditions. In this work, we explore a novel approach that applies the dynamic embedded topic model (DETM) to analyze the evolution of global climate policy discourse. A probabilistic model designed to capture the temporal dynamics of topics over time. We collected a corpus of United Nations Framework Convention on Climate Change (UNFCCC) policy decisions from 1995 to 2023, excluding 2020 due to the postponement of COP26 as a result of the COVID-19 pandemic. The model reveals shifts from early emphases on greenhouse gases and international conventions to recent focuses on implementation, technical collaboration, capacity building, finance, and global agreements. Section 3 presents the modeling pipeline, including preprocessing, model training, and visualization of temporal word distributions. Our results show that DETM is a scalable and effective tool for analyzing the evolution of global policy discourse. Section 4 discusses the implications of these findings and we concluded with future directions and refinements to extend this approach to other policy domains.

Reward modeling (RM), which captures human preferences to align large language models (LLMs), is increasingly employed in tasks such as model finetuning, response filtering, and ranking. However, due to the inherent complexity of human preferences and the limited coverage of available datasets, reward models often fail under distributional shifts or adversarial perturbations. Existing approaches for identifying such failure modes typically rely on prior knowledge about preference distributions or failure attributes, limiting their practicality in real-world settings where such information is unavailable. In this work, we propose a tractable, preference-distribution agnostic method for discovering reward model failure modes via reward guided controlled decoding. Building on this, we introduce REFORM, a self-improving reward modeling framework that enhances robustness by using the reward model itself to guide the generation of falsely scored responses. These adversarial examples are then used to augment the training data and patch the reward model's misaligned behavior. We evaluate REFORM on two widely used preference datasets Anthropic Helpful Harmless (HH) and PKU Beavertails and demonstrate that it significantly improves robustness without sacrificing reward quality. Notably, REFORM preserves performance both in direct evaluation and in downstream policy training, and further improves alignment quality by removing spurious correlations.

The use of Natural Language Processing (NLP) in high-stakes AI-based applications has increased significantly in recent years, especially since the emergence of Large Language Models (LLMs). However, despite their strong performance, LLMs introduce important legal/ethical concerns, particularly regarding privacy, data protection, and transparency. Due to these concerns, this work explores the use of Named-Entity Recognition (NER) to facilitate the privacy-preserving training (or adaptation) of LLMs. We propose a framework that uses NER technologies to anonymize sensitive information in text data, such as personal identities or geographic locations. An evaluation of the proposed privacy-preserving learning framework was conducted to measure its impact on user privacy and system performance in a particular high-stakes and sensitive setup: AI-based resume scoring for recruitment processes. The study involved two language models (BERT and RoBERTa) and six anonymization algorithms (based on Presidio, FLAIR, BERT, and different versions of GPT) applied to a database of 24,000 candidate profiles. The findings indicate that the proposed privacy preservation techniques effectively maintain system performance while playing a critical role in safeguarding candidate confidentiality, thus promoting trust in the experimented scenario. On top of the proposed privacy-preserving approach, we also experiment applying an existing approach that reduces the gender bias in LLMs, thus finally obtaining our proposed Privacy-and Bias-aware LLMs (PBa-LLMs). Note that the proposed PBa-LLMs have been evaluated in a particular setup (resume scoring), but are generally applicable to any other LLM-based AI application.

Lasker, the Times explained, was the "intermediary" who tipped off the publication about Mamdani's application, which was included in a larger hack of Columbia's computer systems. After the Times published its story, Lasker celebrated on X. "I break-uh dah news," he wrote to his more than 260,000 followers. On both X and Substack, where he also has a large following, Lasker is best-known for compiling charts on the "Black-White IQ gap" and otherwise linking race to real-world outcomes. He seems convinced that any differences are the result of biology, and has shot down other possible explanations. He has suggested that crime is genetic.