arXiv subjects

Reihaneh Rabbany

Publications and source records attributed to Reihaneh Rabbany.

At least 19 recordsLinked to original sources

arXiv2026

StabilityBench: Benchmarking Instability in LLMs

AI Assistants are increasingly deployed in high-stakes settings, such as healthcare or government services. Yet their real-world behavior remains poorly understood due to strong context dependence. Current evaluation protocols follow a defense-in-depth paradigm with compounding layers of safeguards, ranging from traditional benchmarks to live or adversarial testing. Such benchmarks remain largely static and single-turn, limiting their ability to capture real-world variability in conversational settings. We propose StabilityBench, a principled, general and model-agnostic benchmark operator that transforms single-turn benchmark queries into multi-turn interaction histories. StabilityBench augments existing benchmarks by injecting realistic user simulations, through demographic proxies or sycophantic baits, while preserving original task intent. We apply StabilityBench to four benchmarks spanning mathematical reasoning, health question-answering and safety, and evaluate nine large language models under these conditions. Our results show that model performance is consistently unstable under these injections, with considerable performance degradations on three out of four benchmarks studied. These highlight important limitations of static evaluations and motivate more realistic evaluation settings. To this end, we propose StabilityBench-Mini: a size-preserving variant of StabilityBench that samples across diversification axes, enabling more realistic evaluation without increasing costs.

cs.LG↗

arXiv2026

EASE Configuration Facilitates A Reproducible Science of LLM Social Simulations

LLMs are increasingly deployed to simulate social interactions, yet many of the existing simulators remain ad hoc and monolithic. This lack of architectural standardization prevents reproducible research and complicates downstream evaluation. We advance a rigorous science of LLM-based multi-agent simulation by modularizing core components into Environments, Agents, Simulation engines, and Evaluation metrics (EASE). We demonstrate the utility of EASE configuration by wrapping it in an experimental study schema for orchestrating workflows centered around answering explicit research questions in generated scenarios. We contribute SiliSocS, an open-source, research-ready Silicon Society Sandbox implementing a study-structured EASE configuration to enable highly configurable and reproducible LLM-based social simulations. Using SiliSocS and EASE, we present three case studies, showcasing the system's comprehensive assessment of existing questions, ability to dive deeper into complex questions, and elaboration of existing studies, respectively. Together, these case studies highlight the limitations of current modeling approaches and isolate the impacts of design choices on key results.

cs.MA↗

arXiv2026

RL Fine-Tuning Heals OOD Forgetting in SFT

Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL) is a standard post-training recipe for improving Large Language Models (LLM) reasoning, but why it works remains unclear. We revisit the common claim that ``SFT memorizes, RL generalizes'' through checkpoint-wise analyses of in-distribution (ID) and out-of-distribution (OOD) reasoning. We find that OOD performance often peaks early during SFT and then declines despite continued improvement in ID reasoning. RL typically does not surpass this early SFT peak; rather, it restores OOD capability lost during later SFT, and only from a bounded range of SFT checkpoints. Further spectral analysis shows that this forgetting-and-recovery pattern correlates with rotations of singular vectors, while singular values remain largely stable. These findings suggest a more precise view of post-training dynamics: SFT can forget, RL can recover, and controlling singular-vector rotation may improve OOD robustness. Code is available at \href{https://github.com/jinhangzhan/RL\_Heals\_SFT.git}{https://github.com/jinhangzhan/RL\_Heals\_SFT}.

cs.LG↗

arXiv2026

Kurtosis-Guided Denoising Score Matching for Tabular Anomaly Detection

Denoising score matching (DSM) provides a way to learn data distributions by training a neural network to recover the score function, defined as the gradient of the log density, from noise-corrupted samples. Once trained, the score magnitude at a test point reflects how consistent that point is with the learned distribution, making it a natural anomaly signal. The key practical challenge is selecting the perturbation scale: too little noise yields unstable score estimates in sparse regions, while too much erases local structure and weakens anomaly sensitivity. This is compounded by the difficulty of hyperparameter tuning when anomalies are unknown and no validation set is available. We introduce kurtosis-based noise scaling (K-DSM), a per-feature scheme that sets noise levels from the shape of each marginal distribution, improving coverage of low-density regions and precision in high-density regions without extra model complexity. Contrary to prior claims that multi-scale or noise-conditioned training is necessary, we find that a carefully trained single-scale model is already a strong anomaly detector. On standard tabular anomaly detection benchmarks, K-DSM achieves state-of-the-art performance in the semi-supervised setting. When combined with a lightweight EMA-teacher filtering rule that removes low-density training points before each gradient step, it also achieves strong performance in the fully unsupervised (contaminated) setting, suggesting that simple, data-adaptive noise scaling enables robust anomaly detection while reducing reliance on hyperparameter tuning.

cs.LG↗

arXiv2026

ControBench: An Interaction-Aware Benchmark for Controversial Discourse Analysis on Social Networks

Understanding how people argue across ideological divides online is important for studying political polarization, misinformation, and content moderation. Existing datasets capture only part of this problem: some preserve text but ignore interaction structure, some model structure without rich semantics, and others represent conversations without stable user-level ideological identity. We introduce ControBench, a benchmark for controversial discourse analysis that combines heterogeneous social interaction graphs with rich textual semantics. Built from Reddit discussions on three topics, Trump, abortion, and religion, ControBench contains 7,370 users, 1,783 posts, and 26,525 interactions. The graph contains user and post nodes connected by semantically enriched edges; in particular, user-comment-user edges encode both a reply and the parent comment that it responds to, preserving local argumentative context. User labels are derived from self-declared Reddit flairs, providing a scalable proxy for ideological identity without manual annotation. The resulting datasets exhibit low or negative adjusted homophily (Trump: -0.77, Abortion: 0.06, Religion: 0.04), reflecting the cross-cutting structure of real-world debate. We evaluate graph neural networks, pretrained language models, and large language models on ControBench and observe distinct performance patterns across topics and model families, especially when ideological boundaries are ambiguous. These results position ControBench as a challenging and realistic benchmark for controversial discourse analysis.

cs.CL↗

arXiv2026

The $\textit{Silicon Society}$ Cookbook: Design Space of LLM-based Social Simulations

Studies attempting to simulate human behavior with $\textit{Silicon Societies}$ grow in numbers while LLM-only social networks have started appearing outside of controlled settings. However, the design space of these networks remains under-studied, which contributes to a gap in validating model realism. To enable future works to make more informed design decisions, we perform a systematic analysis of the consequences and interactions of key design choices in simulated social networks, including the choice of base model used to model individual agents, and how they are connected to each other. Using surveys as a proxy for agent opinions, our findings suggest that the geometry of the design space is non-trivial, with some parameters behaving in additive ways while others display more complex interactions. In particular, the choice of the base LLM is the most important variable impacting the simulation outcomes.

cs.MA↗

arXiv2026

What do people want to fact-check?

Research on misinformation has focused almost exclusively on supply, asking what falsehoods circulate, who produces them, and whether corrections work. A basic demand-side question remains unanswered. When ordinary people can fact-check anything they want, what do they actually ask about? We provide the first large-scale evidence on this question by analyzing close to 2{,}500 statements submitted by 457 participants to an open-ended AI fact-checking system. Each claim is classified along five semantic dimensions (domain, epistemic form, verifiability, target entity, and temporal reference), producing a behavioral map of public verification demand. Three findings stand out. First, users range widely across topics but default to a narrow epistemic repertoire, overwhelmingly submitting simple descriptive claims about present-day observables. Second, roughly one in four requests concerns statements that cannot be empirically resolved, including moral judgments, speculative predictions, and subjective evaluations, revealing a systematic mismatch between what users seek from fact-checking tools and what such tools can deliver. Third, comparison with the FEVER benchmark dataset exposes sharp structural divergences across all five dimensions, indicating that standard evaluation corpora encode a synthetic claim environment that does not resemble real-world verification needs. These results reframe fact-checking as a demand-driven problem and identify where current AI systems and benchmarks are misaligned with the uncertainty people actually experience.

cs.HC↗

arXiv2025

Deepfakes in the 2025 Canadian Election: Prevalence, Partisanship, and Platform Dynamics

Concerns about AI-generated political content are growing, yet there is limited empirical evidence on how deepfakes actually appear and circulate across social platforms during major events in democratic countries. In this study, we present one of the first in-depth analyses of how these realistic synthetic media shape the political landscape online, focusing specifically on the 2025 Canadian federal election. By analyzing 187,778 posts from X, Bluesky, and Reddit with a high-accuracy detection framework trained on a diverse set of modern generative models, we find that 5.86% of election-related images were deepfakes. Right-leaning accounts shared them more frequently, with 8.66% of their posted images flagged compared to 4.42% for left-leaning users, often with defamatory or conspiratorial intent. Yet, most detected deepfakes were benign or non-political, and harmful ones drew little attention, accounting for only 0.12% of all views on X. Overall, deepfakes were present in the election conversation, but their reach was modest, and realistic fabricated images, although less common, drew higher engagement, highlighting growing concerns about their potential misuse.

cs.SI↗

arXiv2025

Higher-Order Transformers With Kronecker-Structured Attention

Modern datasets are increasingly high-dimensional and multiway, often represented as tensor-valued data with multi-indexed variables. While Transformers excel in sequence modeling and high-dimensional tasks, their direct application to multiway data is computationally prohibitive due to the quadratic cost of dot-product attention and the need to flatten inputs, which disrupts tensor structure and cross-dimensional dependencies. We propose the Higher-Order Transformer (HOT), a novel factorized attention framework that represents multiway attention as sums of Kronecker products or sums of mode-wise attention matrices. HOT efficiently captures dense and sparse relationships across dimensions while preserving tensor structure. Theoretically, HOT retains the expressiveness of full high-order attention and allows complexity control via factorization rank. Experiments on 2D and 3D datasets show that HOT achieves competitive performance in multivariate time series forecasting and image classification, with significantly reduced computational and memory costs. Visualizations of mode-wise attention matrices further reveal interpretable high-order dependencies learned by HOT, demonstrating its versatility for complex multiway data across diverse domains. The implementation of our proposed method is publicly available at https://github.com/s-omranpour/HOT.

cs.LG↗

arXiv2025

Grounding Computer Use Agents on Human Demonstrations

Building reliable computer-use agents requires grounding: accurately connecting natural language instructions to the correct on-screen elements. While large datasets exist for web and mobile interactions, high-quality resources for desktop environments are limited. To address this gap, we introduce GroundCUA, a large-scale desktop grounding dataset built from expert human demonstrations. It covers 87 applications across 12 categories and includes 56K screenshots, with every on-screen element carefully annotated for a total of over 3.56M human-verified annotations. From these demonstrations, we generate diverse instructions that capture a wide range of real-world tasks, providing high-quality data for model training. Using GroundCUA, we develop the GroundNext family of models that map instructions to their target UI elements. At both 3B and 7B scales, GroundNext achieves state-of-the-art results across five benchmarks using supervised fine-tuning, while requiring less than one-tenth the training data of prior work. Reinforcement learning post-training further improves performance, and when evaluated in an agentic setting on the OSWorld benchmark using o3 as planner, GroundNext attains comparable or superior results to models trained with substantially more data,. These results demonstrate the critical role of high-quality, expert-driven datasets in advancing general-purpose computer-use agents.

cs.LG↗

arXiv2025

TGM: a Modular and Efficient Library for Machine Learning on Temporal Graphs

Well-designed open-source software drives progress in Machine Learning (ML) research. While static graph ML enjoys mature frameworks like PyTorch Geometric and DGL, ML for temporal graphs (TG), networks that evolve over time, lacks comparable infrastructure. Existing TG libraries are often tailored to specific architectures, hindering support for diverse models in this rapidly evolving field. Additionally, the divide between continuous- and discrete-time dynamic graph methods (CTDG and DTDG) limits direct comparisons and idea transfer. To address these gaps, we introduce Temporal Graph Modelling (TGM), a research-oriented library for ML on temporal graphs, the first to unify CTDG and DTDG approaches. TGM offers first-class support for dynamic node features, time-granularity conversions, and native handling of link-, node-, and graph-level tasks. Empirically, TGM achieves an average 7.8x speedup across multiple models, datasets, and tasks compared to the widely used DyGLib, and an average 175x speedup on graph discretization relative to available implementations. Beyond efficiency, we show in our experiments how TGM unlocks entirely new research possibilities by enabling dynamic graph property prediction and time-driven training paradigms, opening the door to questions previously impractical to study. TGM is available at https://github.com/tgm-team/tgm

cs.LG↗

arXiv2025

OpenFake: An Open Dataset and Platform Toward Real-World Deepfake Detection

Deepfakes, synthetic media created using advanced AI techniques, pose a growing threat to information integrity, particularly in politically sensitive contexts. This challenge is amplified by the increasing realism of modern generative models, which our human perception study confirms are often indistinguishable from real images. Yet, existing deepfake detection benchmarks rely on outdated generators or narrowly scoped datasets (e.g., single-face imagery), limiting their utility for real-world detection. To address these gaps, we present OpenFake, a large politically grounded dataset specifically crafted for benchmarking against modern generative models with high realism, and designed to remain extensible through an innovative crowdsourced adversarial platform that continually integrates new hard examples. OpenFake comprises nearly four million total images: three million real images paired with descriptive captions and almost one million synthetic counterparts from state-of-the-art proprietary and open-source models. Detectors trained on OpenFake achieve near-perfect in-distribution performance, strong generalization to unseen generators, and high accuracy on a curated in-the-wild social media test set, significantly outperforming models trained on existing datasets. Overall, we demonstrate that with high-quality and continually updated benchmarks, automatic deepfake detection is both feasible and effective in real-world settings.

cs.CV↗

arXiv2025

Hallucination Detox: Sensitivity Dropout (SenD) for Large Language Model Training

As large language models (LLMs) become increasingly prevalent, concerns about their reliability, particularly due to hallucinations - factually inaccurate or irrelevant outputs - have grown. Our research investigates the relationship between the uncertainty in training dynamics and the emergence of hallucinations. Using models from the Pythia suite and several hallucination detection metrics, we analyze hallucination trends and identify significant variance during training. To address this, we propose Sensitivity Dropout (SenD), a novel training protocol designed to reduce hallucination variance during training by deterministically dropping embedding indices with significant variability. In addition, we develop an unsupervised hallucination detection metric, Efficient EigenScore (EES), which approximates the traditional EigenScore in 2x speed. This metric is integrated into our training protocol, allowing SenD to be both computationally scalable and effective at reducing hallucination variance. SenD improves test-time reliability of Pythia and Meta's Llama models by up to 17% and enhances factual accuracy in Wikipedia, Medical, Legal, and Coding domains without affecting downstream task performance.

cs.AI↗

arXiv2025

CrediBench: Building Web-Scale Network Datasets for Information Integrity

Automatically assessing the credibility of online sources presents an invaluable tool for navigating today's information ecosystem. However, existing approaches either depend on scarce and costly human annotations, or focus exclusively on assessments at the level of individual claims. Misinformation often spreads via interlinked web domains, whose connections evolve over time. Focusing on claims alone ignores these structural and temporal credibility signals evident in the changing web topology. Existing datasets fail to capture these central modalities in web domain credibility prediction: namely, internet topology, temporality and text (webpage) content. To address this gap, we present CrediBench, a dataset containing eights months of web graph data; of which we analyze the three months surrounding the 2024 U.S. federal elections, a time of heightened misinformation propagation online. Each monthly snapshot contains over 40 million nodes, their scraped webpage content, and over 1 billion hyperlink edges. CrediBench supports credibility prediction as both a regression (continuous credibility score) and a binary classification task (credible or not). For classification, we curate a novel binary label set containing 662,575 web domains labelled for boolean credibility, spanning four areas (misinformation, crowd-sourced, malware and phishing). Our empirical experiments support that all task modalities-graph, text and time-contribute significantly to achieving the best performance. In particular, our multi-modal regression model trained on CrediBench outperforms other configurations and existing baselines, decreasing Mean Average Error from 0.162 to 0.107 on the regression task, while the multi-modal classifier improves accuracy from 56% to 85% on the classification one. CrediBench, our proposed web-scale multi-modal dataset, is available on Huggingface for future research.

cs.SI↗

arXiv2025

$\texttt{BluePrint}$: A Social Media User Dataset for LLM Persona Evaluation and Training

Large language models (LLMs) offer promising capabilities for simulating social media dynamics at scale, enabling studies that would be ethically or logistically challenging with human subjects. However, the field lacks standardized data resources for fine-tuning and evaluating LLMs as realistic social media agents. We address this gap by introducing SIMPACT, the SIMulation-oriented Persona and Action Capture Toolkit, a privacy respecting framework for constructing behaviorally-grounded social media datasets suitable for training agent models. We formulate next-action prediction as a task for training and evaluating LLM-based agents and introduce metrics at both the cluster and population levels to assess behavioral fidelity and stylistic realism. As a concrete implementation, we release BluePrint, a large-scale dataset built from public Bluesky data focused on political discourse. BluePrint clusters anonymized users into personas of aggregated behaviours, capturing authentic engagement patterns while safeguarding privacy through pseudonymization and removal of personally identifiable information. The dataset includes a sizable action set of 12 social media interaction types (likes, replies, reposts, etc.), each instance tied to the posting activity preceding it. This supports the development of agents that use context-dependence, not only in the language, but also in the interaction behaviours of social media to model social media users. By standardizing data and evaluation protocols, SIMPACT provides a foundation for advancing rigorous, ethically responsible social media simulations. BluePrint serves as both an evaluation benchmark for political discourse modeling and a template for building domain specific datasets to study challenges such as misinformation and polarization.

cs.CL↗

arXiv2025

A Guide to Misinformation Detection Data and Evaluation

Misinformation is a complex societal issue, and mitigating solutions are difficult to create due to data deficiencies. To address this, we have curated the largest collection of (mis)information datasets in the literature, totaling 75. From these, we evaluated the quality of 36 datasets that consist of statements or claims, as well as the 9 datasets that consist of data in purely paragraph form. We assess these datasets to identify those with solid foundations for empirical work and those with flaws that could result in misleading and non-generalizable results, such as spurious correlations, or examples that are ambiguous or otherwise impossible to assess for veracity. We find the latter issue is particularly severe and affects most datasets in the literature. We further provide state-of-the-art baselines on all these datasets, but show that regardless of label quality, categorical labels may no longer give an accurate evaluation of detection model performance. Finally, we propose and highlight Evaluation Quality Assurance (EQA) as a tool to guide the field toward systemic solutions rather than inadvertently propagating issues in evaluation. Overall, this guide aims to provide a roadmap for higher quality data and better grounded evaluations, ultimately improving research in misinformation detection. All datasets and other artifacts are available at https://misinfo-datasets.complexdatalab.com/.

cs.SI↗

arXiv2025

The Singapore Consensus on Global AI Safety Research Priorities

Rapidly improving AI capabilities and autonomy hold significant promise of transformation, but are also driving vigorous debate on how to ensure that AI is safe, i.e., trustworthy, reliable, and secure. Building a trusted ecosystem is therefore essential -- it helps people embrace AI with confidence and gives maximal space for innovation while avoiding backlash. The "2025 Singapore Conference on AI (SCAI): International Scientific Exchange on AI Safety" aimed to support research in this space by bringing together AI scientists across geographies to identify and synthesise research priorities in AI safety. This resulting report builds on the International AI Safety Report chaired by Yoshua Bengio and backed by 33 governments. By adopting a defence-in-depth model, this report organises AI safety research domains into three types: challenges with creating trustworthy AI systems (Development), challenges with evaluating their risks (Assessment), and challenges with monitoring and intervening after deployment (Control).

cs.AI↗

arXiv2025

Ask before you Build: Rethinking AI-for-Good in Human Trafficking Interventions

AI for good initiatives often rely on the assumption that technical interventions can resolve complex social problems. In the context of human trafficking (HT), such techno-solutionism risks oversimplifying exploitation, reinforcing power imbalances and causing harm to the very communities AI claims to support. In this paper, we introduce the Radical Questioning (RQ) framework as a five step, pre-project ethical assessment tool to critically evaluate whether AI should be built at all, especially in domains involving marginalized populations and entrenched systemic injustice. RQ does not replace principles based ethics but precedes it, offering an upstream, deliberative space to confront assumptions, map power, and consider harms before design. Using a case study in AI for HT, we demonstrate how RQ reveals overlooked sociocultural complexities and guides us away from surveillance based interventions toward survivor empowerment tools. While developed in the context of HT, RQ's five step structure can generalize to other domains, though the specific questions must be contextual. This paper situates RQ within a broader AI ethics philosophy that challenges instrumentalist norms and centers relational, reflexive responsibility.

cs.CY↗