Cautious predictions -- where a machine learning model abstains when uncertain -- are crucial for limiting harmful errors in safety-critical applications. In this work, we identify a novel threat: a dishonest institution can exploit these mechanisms to discriminate or unjustly deny services under the guise of uncertainty. We demonstrate the practicality of this threat by introducing an uncertainty-inducing attack called Mirage, which deliberately reduces confidence in targeted input regions, thereby covertly disadvantaging specific individuals. At the same time, Mirage maintains high predictive performance across all data points. To counter this threat, we propose Confidential Guardian, a framework that analyzes calibration metrics on a reference dataset to detect artificially suppressed confidence. Additionally, it employs zero-knowledge proofs of verified inference to ensure that reported confidence scores genuinely originate from the deployed model. This prevents the provider from fabricating arbitrary model confidence values while protecting the model's proprietary details. Our results confirm that Confidential Guardian effectively prevents the misuse of cautious predictions, providing verifiable assurances that abstention reflects genuine model uncertainty rather than malicious intent.

Human-AI conversation frequently relies on quoting earlier text-"check it with the formula I just highlighted"-yet today's large language models (LLMs) lack an explicit mechanism for locating and exploiting such spans. We formalise the challenge as span-conditioned generation, decomposing each turn into the dialogue history, a set of token-offset quotation spans, and an intent utterance. Building on this abstraction, we introduce a quotation-centric data pipeline that automatically synthesises task-specific dialogues, verifies answer correctness through multi-stage consistency checks, and yields both a heterogeneous training corpus and the first benchmark covering five representative scenarios. To meet the benchmark's zero-overhead and parameter-efficiency requirements, we propose QuAda, a lightweight training-based method that attaches two bottleneck projections to every attention head, dynamically amplifying or suppressing attention to quoted spans at inference time while leaving the prompt unchanged and updating < 2.8% of backbone weights. Experiments across models show that QuAda is suitable for all scenarios and generalises to unseen topics, offering an effective, plug-and-play solution for quotation-aware dialogue.

The increasing use of LLMs as substitutes for humans in ``aligning'' LLMs has raised questions about their ability to replicate human judgments and preferences, especially in ambivalent scenarios where humans disagree. This study examines the biases and limitations of LLMs in three roles: answer generator, judge, and debater. These roles loosely correspond to previously described alignment frameworks: preference alignment (judge) and scalable oversight (debater), with the answer generator reflecting the typical setting with user interactions. We develop a ``no-consensus'' benchmark by curating examples that encompass a variety of a priori ambivalent scenarios, each presenting two possible stances. Our results show that while LLMs can provide nuanced assessments when generating open-ended answers, they tend to take a stance on no-consensus topics when employed as judges or debaters. These findings underscore the necessity for more sophisticated methods for aligning LLMs without human oversight, highlighting that LLMs cannot fully capture human disagreement even on topics where humans themselves are divided.

The rapid evolution of social media has generated an overwhelming volume of user-generated content, conveying implicit opinions and contributing to the spread of misinformation. The method aims to enhance the detection of stance where misinformation can polarize user opinions. Stance detection has emerged as a crucial approach to effectively analyze underlying biases in shared information and combating misinformation. This paper proposes a novel method for \textbf{S}tance \textbf{P}rediction through a \textbf{L}abel-fused dual cross-\textbf{A}ttentive \textbf{E}motion-aware neural \textbf{Net}work (SPLAENet) in misinformative social media user-generated content. The proposed method employs a dual cross-attention mechanism and a hierarchical attention network to capture inter and intra-relationships by focusing on the relevant parts of source text in the context of reply text and vice versa. We incorporate emotions to effectively distinguish between different stance categories by leveraging the emotional alignment or divergence between the texts. We also employ label fusion that uses distance-metric learning to align extracted features with stance labels, improving the method's ability to accurately distinguish between stances. Extensive experiments demonstrate the significant improvements achieved by SPLAENet over existing state-of-the-art methods. SPLAENet demonstrates an average gain of 8.92\% in accuracy and 17.36\% in F1-score on the RumourEval dataset. On the SemEval dataset, it achieves average gains of 7.02\% in accuracy and 10.92\% in F1-score. On the P-stance dataset, it demonstrates average gains of 10.03\% in accuracy and 11.18\% in F1-score. These results validate the effectiveness of the proposed method for stance detection in the context of misinformative social media content.

This paper focuses on Zero-Trust Foundation Models (ZTFMs), a novel paradigm that embeds zero-trust security principles into the lifecycle of foundation models (FMs) for Internet of Things (IoT) systems. By integrating core tenets, such as continuous verification, least privilege access (LPA), data confidentiality, and behavioral analytics into the design, training, and deployment of FMs, ZTFMs can enable secure, privacy-preserving AI across distributed, heterogeneous, and potentially adversarial IoT environments. We present the first structured synthesis of ZTFMs, identifying their potential to transform conventional trust-based IoT architectures into resilient, self-defending ecosystems. Moreover, we propose a comprehensive technical framework, incorporating federated learning (FL), blockchain-based identity management, micro-segmentation, and trusted execution environments (TEEs) to support decentralized, verifiable intelligence at the network edge. In addition, we investigate emerging security threats unique to ZTFM-enabled systems and evaluate countermeasures, such as anomaly detection, adversarial training, and secure aggregation. Through this analysis, we highlight key open research challenges in terms of scalability, secure orchestration, interpretable threat attribution, and dynamic trust calibration. This survey lays a foundational roadmap for secure, intelligent, and trustworthy IoT infrastructures powered by FMs.

The rapid rise of Generative AI (GenAI) tools has sparked debate over their role in complementing or replacing human workers across job contexts. We present a mathematical framework that models jobs, workers, and worker-job fit, introducing a novel decomposition of skills into decision-level and action-level subskills to reflect the complementary strengths of humans and GenAI. We analyze how changes in subskill abilities affect job success, identifying conditions for sharp transitions in success probability. We also establish sufficient conditions under which combining workers with complementary subskills significantly outperforms relying on a single worker. This explains phenomena such as productivity compression, where GenAI assistance yields larger gains for lower-skilled workers. We demonstrate the framework' s practicality using data from O*NET and Big-Bench Lite, aligning real-world data with our model via subskill-division methods. Our results highlight when and how GenAI complements human skills, rather than replacing them.

Training data mixtures greatly impact the generalization performance of large language models. Existing domain reweighting methods often rely on costly weight computations and require retraining when new data is introduced. To this end, we introduce a flexible and efficient data mixing framework, Chameleon, that employs leverage scores to quantify domain importance within a learned embedding space. We first construct a domain affinity matrix over domain embeddings. The induced leverage scores determine a mixture that upweights domains sharing common representations in embedding space. This formulation allows direct transfer to new data by computing the new domain embeddings. In experiments, we demonstrate improvements over three key scenarios: (i) our computed weights improve performance on pretraining domains with a fraction of the compute of existing methods; (ii) Chameleon can adapt to data changes without proxy retraining, boosting few-shot reasoning accuracies when transferred to new data; (iii) our method enables efficient domain reweighting in finetuning, consistently improving test perplexity on all finetuning domains over uniform mixture. Our code is available at https://github.com/LIONS-EPFL/Chameleon.

The widespread success of foundation models in natural language processing and computer vision has inspired researchers to extend the concept to scientific machine learning and computational science. However, this position paper argues that as the term "foundation model" is an evolving concept, its application in computational science is increasingly used without a universally accepted definition, potentially creating confusion and diluting its precise scientific meaning. In this paper, we address this gap by proposing a formal definition of foundation models in computational science, grounded in the core values of generality, reusability, and scalability. We articulate a set of essential and desirable characteristics that such models must exhibit, drawing parallels with traditional foundational methods, like the finite element and finite volume methods. Furthermore, we introduce the Data-Driven Finite Element Method (DD-FEM), a framework that fuses the modular structure of classical FEM with the representational power of data-driven learning. We demonstrate how DD-FEM addresses many of the key challenges in realizing foundation models for computational science, including scalability, adaptability, and physics consistency. By bridging traditional numerical methods with modern AI paradigms, this work provides a rigorous foundation for evaluating and developing novel approaches toward future foundation models in computational science.

Despite well-documented consequences of the U.S. government's 1930s housing policies on racial wealth disparities, scholars have struggled to quantify its precise financial effects due to the inaccessibility of historical property appraisal records. Many counties still store these records in physical formats, making large-scale quantitative analysis difficult. We present an approach scholars can use to digitize historical housing assessment data, applying it to build and release a dataset for one county. Starting from publicly available scanned documents, we manually annotated property cards for over 12,000 properties to train and validate our methods. We use OCR to label data for an additional 50,000 properties, based on our two-stage approach combining classical computer vision techniques with deep learning-based OCR. For cases where OCR cannot be applied, such as when scanned documents are not available, we show how a regression model based on building feature data can estimate the historical values, and test the generalizability of this model to other counties. With these cost-effective tools, scholars, community activists, and policy makers can better analyze and understand the historical impacts of redlining.

Large Language Models (LLMs) excel in various natural language processing tasks but remain vulnerable to generating harmful content or being exploited for malicious purposes. Although safety alignment datasets have been introduced to mitigate such risks through supervised fine-tuning (SFT), these datasets often lack comprehensive risk coverage. Most existing datasets focus primarily on lexical diversity while neglecting other critical dimensions. To address this limitation, we propose a novel analysis framework to systematically measure the risk coverage of alignment datasets across three essential dimensions: Lexical Diversity, Malicious Intent, and Jailbreak Tactics. We further introduce TRIDENT, an automated pipeline that leverages persona-based, zero-shot LLM generation to produce diverse and comprehensive instructions spanning these dimensions. Each harmful instruction is paired with an ethically aligned response, resulting in two datasets: TRIDENT-Core, comprising 26,311 examples, and TRIDENT-Edge, with 18,773 examples. Fine-tuning Llama 3.1-8B on TRIDENT-Edge demonstrates substantial improvements, achieving an average 14.29% reduction in Harm Score, and a 20% decrease in Attack Success Rate compared to the best-performing baseline model fine-tuned on the WildBreak dataset.