Mental manipulation on social media poses a covert yet serious threat to individuals' psychological well-being and the integrity of online interactions. Detecting such behavior is challenging due to the difficult-to-annotate training data, its highly covert and multi-turn nature, and the lack of real-world datasets. To address these challenges, we propose MentalMAD, a framework that enhances large language models for mental manipulation detection. Our approach consists of three key components: EvoSA, an annotation-free data augmentation method that combines evolutionary operations with speech-act-aware prompting; teacher-model-generated complementary-task supervision; and Complementary-Convergent Distillation, a phase-wise strategy for transferring manipulation-specific knowledge to student models. We then constructed the ReaMent dataset, comprising 5,000 real-world-sourced dialogues. Extensive experiments show that MentalMAD improves accuracy by 14.0%, macro-F1 by 27.3%, and weighted F1 by 15.1% over the strongest baseline. The code and the dataset are publicly available at https://github.com/Yuansheng-Gao/MentalMAD.
论文检索
输入标题、作者或关键词，从 5,826 篇学术成果中精准定位
The rapid growth of video content on platforms such as TikTok and YouTube has intensified the spread of multimodal hate speech, where harmful cues emerge subtly and asynchronously across visual, acoustic, and textual streams. Existing research primarily focuses on video-level classification, leaving the practically crucial task of temporal localisation, identifying when hateful segments occur, largely unaddressed. This challenge is even more noticeable under weak supervision, where only video-level labels are available, and static fusion or classification-based architectures struggle to capture cross-modal and temporal dynamics. To address these challenges, we propose MultiHateLoc, the first framework designed for weakly-supervised multimodal hate localisation. MultiHateLoc incorporates (1) modality-aware temporal encoders to model heterogeneous sequential patterns, including a tailored text-based preprocessing module for feature enhancement; (2) dynamic cross-modal fusion to adaptively emphasise the most informative modality at each moment and a cross- modal contrastive alignment strategy to enhance multimodal feature consistency; (3) a modality-aware MIL objective to identify discriminative segments under video-level supervision. Despite relying solely on coarse labels, MultiHateLoc produces fine-grained, interpretable frame-level predictions. Experiments on HateMM and MultiHateClip show that our method achieves state-of-the-art performance in the localisation task. Code is available at https://github.com/Multimodal-Intelligence-Lab-MIL/MultiHateLoc.
Online financial services constitute an essential component of contemporary web ecosystems, yet their openness introduces substantial exposure to fraud that harms vulnerable users and weakens trust in digital finance. Such threats have become a significant web harm that erodes societal fairness and affects the well-being of online communities. However, existing detection methods based on graph neural networks (GNNs) struggle with two persistent challenges: (1) long-tailed data distributions, which obscure rare but critical fraudulent cases, and (2) fraud camouflage, where malicious transactions mimic benign behaviors to evade detection. To fill these gaps, we propose HIMVH, a Hippocampus-Inspired Multi-View Hypergraph learning model for web finance fraud detection. Specifically, drawing inspiration from the scene conflict monitoring role of the hippocampus, we design a cross-view inconsistency perception module that captures subtle discrepancies and behavioral heterogeneity across multiple transaction views. This module enables the model to identify subtle cross-view conflicts for detecting online camouflaged fraudulent behaviors. Furthermore, inspired by the match-mismatch novelty detection mechanism of the CA1 region, we introduce a novelty-aware hypergraph learning module that measures feature deviations from neighborhood expectations and adaptively reweights messages, thereby enhancing sensitivity to online rare fraud patterns in the long-tailed settings. Extensive experiments on six web-based financial fraud datasets demonstrate that HIMVH achieves 6.42% improvement in AUC, 9.74% in F1 and 39.14% in AP on average over 15 SOTA models.
Understanding experienced segregation in urban daily life is crucial for addressing societal inequalities and fostering inclusivity. The abundance of user-generated reviews on social media encapsulates nuanced perceptions and feelings associated with different places, offering rich insights into segregation. However, leveraging this data poses significant challenges due to its vast volume, ambiguity, and confluence of diverse perspectives. To tackle these challenges, we propose a novel Large Language Model (LLM) Agent to automate online review mining for segregation prediction. Specifically, we propose a reflective LLM coder to digest social media content into insights consistent with real-world feedback, and eventually produce a codebook capturing key dimensions that signal segregation experience, such as cultural resonance and appeal, accessibility and convenience, and community engagement and local involvement. Guided by the codebook, LLMs can generate both informative review summaries and ratings for segregation prediction. Moreover, we design a REasoning-and-EMbedding (RE'EM) framework, which combines the reasoning and embedding capabilities of language models to integrate multi-channel features for segregation prediction. Experiments on real-world data demonstrate that our agent substantially improves prediction accuracy, with a 22.79% elevation in R2 and a 9.33% reduction in MSE. The derived codebook is generalizable across three different cities, consistently improving prediction accuracy. Moreover, our user study confirms that the codebook-guided summaries provide cognitive gains for human participants in perceiving places of interest(POIs)' social inclusiveness. Our study marks an important step toward understanding implicit social barriers and inequalities, demonstrating the great potential of promoting social inclusiveness with Web technology.
Community-based fact-checking systems, such as Community Notes on X (formerly Twitter), aim to mitigate online misinformation by surfacing annotations judged helpful by contributors with diverse viewpoints. While prior work has shown that the platform's bridging-based algorithm effectively selects helpful notes at the time of display, little is known about how evaluations change after notes become visible. Using a large-scale dataset of 437,396 community notes and 35 million ratings from over 580,000 contributors, we examine the stability of helpful notes and the rating dynamics that follow their initial display. We find that 30.2% of displayed notes later lose their helpful status and disappear. Using interrupted time series models, we further show that note display triggers a sharp increase in rating volume and a significant shift in rating leaning, but these effects differ across rater groups. Contributors with viewpoints similar to note authors tend to increase supportive ratings, while dissimilar contributors increase negative ratings, producing systematic post-display polarization. Counterfactual analyses suggest that this post-display polarization, particularly from dissimilar raters, plays a substantial role in note disappearance. These findings highlight the vulnerability of consensus-based fact-checking systems to polarized rating behavior and suggest pathways for improving their resilience.
As Large Language Models (LLMs) increasingly shape online content, how to remove targeted information from well-trained LLMs (also known as LLM unlearning) has become increasingly critical for web governance. A key challenge in LLM unlearning lies in the sample-wise imbalance within the forget set: different samples exhibit widely varying unlearning difficulty, leading to asynchronous forgetting speeds where some knowledge remains insufficiently erased while others become over-forgotten. To address this challenge, we propose BalDRO, a novel and efficient framework for balanced LLM unlearning. BalDRO formulates unlearning as a min–sup process, where the inner process identifies a worst-case data distribution that adaptively emphasizes hard-to-unlearn samples, while the outer process updates model parameters based on the worst-case data distribution. We instantiate this formulation through two efficient variants: BalDRO-G, a discrete GroupDRO-based approximation that focuses on high-loss subsets, and BalDRO-DV, a continuous Donsker–Varadhan dual method that enables smooth, adaptive weighting within standard LLM training pipelines. Extensive experiments on the TOFU and MUSE benchmarks demonstrate the effectiveness of our proposed BalDRO, yielding significant improvements in both forgetting quality and model utility over existing methods. For reproducibility, we have released the code for BalDRO. https://github.com/nxZhai/BalDRO.
Online advertising platforms serve as a critical bridge between advertisers and media, requiring precise prediction of user behaviors. Systematic underestimation or overestimation in these predictions can undermine the interests of both parties. Existing calibration methods fall short in addressing two key challenges. First, significant differences in CVR distributions across various targets lead to biased calibration when using global posterior statistics, causing some sample groups to be overestimated while others are underestimated. Second, current field-aware approaches are typically limited to single-field calibration and fail to account for field sensitivity. To overcome these limitations, we propose a Pareto Frontier-based Multi-field Personalized Calibration (PF-MPC) method which formulates multi-field calibration as a multi-objective optimization problem. PF-MPC identifies the optimal Pareto-efficient weight combinations to balance the conflicting calibration errors across different fields. We evaluate PF-MPC on both public calibration benchmark and a large-scale industrial dataset. Experimental results demonstrate that our method achieves significant improvements in calibration performance compared to existing approaches.
User behavior on social media---from scrolling and viewing to liking, reposting, and posting---yet most research relies on self-reports that obscure fine-grained usage patterns. We analyze high-resolution activity logs from 209 U.S. X (Twitter) users tracked over four weeks to identify distinct behavioral profiles based on session-level features. Latent profile analysis reveals three groups---interactors (32.52%), lurkers (60.45%), and creators (7.03%) that differ in engagement intensity, demographics, and content exposure. Interactors and lurkers skew younger and Democratic, whereas creators skew older and more Republican, consuming more ideological and low-credibility content. These results link behavioral heterogeneity to systematically different information environments and suggest that platform interventions may operate unevenly across user types. Our findings demonstrate the value of log-based behavioral segmentation for understanding online participation and motivate profile-aware platform governance and content moderation strategies.
Trending news detection in low-traffic search environments faces a fundamental cold-start problem, where a lack of query volume prevents systems from identifying emerging or long-tail trends. Existing methods relying on keyword frequency or query spikes are inherently slow and ineffective in these sparse settings, lagging behind real-world shifts in attention. We introduce RTTP, a novel Real-Time Trending Prediction framework that generates search queries directly from news content instead of waiting for users to issue them. RTTP leverages a continual learning LLM (CL-LLM) that converts posts into search-style queries and scores them using engagement strength + creator authority, enabling early trend surfacing before search volume forms. To ensure adaptation without degrading reasoning, we propose Mix-Policy DPO, a new preference-based continual learning approach that combines on-policy stability with off-policy novelty to mitigate catastrophic forgetting during model upgrades. Deployed at production scale on Facebook and Meta AI products, RTTP delivers +91.4% improvement in tail-trend detection precision@500 and +19% query generation accuracy over industry baselines, while sustaining stable performance after multi-week online training. This work demonstrates that LLM-generated synthetic search signals, when aligned and continually updated, unlock timely trend understanding in low-traffic search environments.
In this paper, we propose Pr ecision-Informed Semantic Modeling (PRISM), a structured topic modeling framework combining the benefits of rich representations captured by LLMs with the low cost and interpretability of latent semantic clustering methods. PRISM fine-tunes a sentence encoding model using a sparse set of LLM-provided labels on samples drawn from some corpus of interest. We segment this embedding space with thresholded clustering, yielding clusters that separate closely related topics within some narrow domain. Across multiple corpora, PRISM improves topic separability over state-of-the-art local topic models and even over clustering on large, frontier embedding models while requiring only a small number of LLM queries to train. This work contributes to several research streams by providing (i) a student-teacher pipeline to distill sparse LLM supervision into a lightweight model for topic discovery; (ii) an analysis of the efficacy of sampling strategies to improve local geometry for cluster separability; and (iii)~an effective approach for web-scale text analysis, enabling researchers and practitioners to track nuanced claims and subtopics online with an interpretable, locally deployable framework.
Online marketplaces will be transformed by autonomous AI agents acting on behalf of consumers. Rather than humans browsing and clicking, AI agents can parse webpages or interact through APIs to evaluate products, and transact. This raises a fundamental question: what do AI agents buy—and why? We develop ACES, a sandbox environment that pairs a platform-agnostic agent with a fully programmable mock marketplace to study this. We first explore aggregate choices, revealing that modal choices can differ across models, with AI agents sometimes concentrating on a few products, raising competition questions. We then analyze the current drivers of choices through randomized experiments on product positions and listing attributes. Models show sizeable and heterogeneous position effects: all favor the top row, yet different models prefer different columns, undermining the assumption of a universal ''top'' rank. They penalize sponsored tags, reward endorsements, and sensitivities to price, ratings, and reviews are directionally as expected, but vary sharply across models. Our findings reveal how AI agents behave in e-commerce, and surface concrete monitoring, seller strategy, platform design, and regulatory questions.
Traditional vertical federated learning schema suffers from two main issues: 1) restricted applicable scope to overlapped samples and 2) high system challenge of real-time federated serving, which limits its application to advertising systems. To this end, we advocate a new practical learning setting, Semi-VFL (Vertical Semi-Federated Learning), for real-world industrial applications, where the learned model retains sufficient advantages of federated learning while supporting independent local serving. To achieve this goal, we propose the carefully designed Joint Privileged Learning framework (JPL ) to i) alleviate the absence of the passive party's feature with federated equivalence imitation and ii) adapt to the heterogeneous full sample space with cross-branch rank alignment. Extensive experiments conducted on real-world advertising datasets validate the effectiveness of our method over baseline methods.
In large-scale e-commerce platforms, Conversion Rate (CVR) prediction is crucial for recommender system, yet existing approach face a fundamental granularity mismatch: models operate at the item level while users purchase at the fine-grained Stock Keeping Unit (SKU) level. This mismatch causes loss of fine-grained user intent signals. Moreover, it also introduces price inconsistency bias due to the gap between static exposure prices and actual transaction prices. While direct SKU-level modeling would resolve these issues, it is impractical for industrial deployment due to the extreme data sparsity and prohibitive inference costs. To address these challenges, we propose PSKU4Rec, a novel framework that operates at the Price SKU (PSKU) granularity by aggregating SKUs by price to retain critical price signals while reducing sparsity by 6 times. PSKU4Rec consists of: (1) a PSKU-aware prediction network that models intra-item PSKU contextual information and captures user PSKU preferences; and (2) a PSKU-aware application module that generates personalized estimated transaction prices for pCVR refinement and enables personalized main-image display. Offline experiments based on the dataset collected from Taobao App show substantial improvements in CVR prediction accuracy and price consistency. Online A/B testing further validates the effectiveness of our approach.
Health misinformation poses serious public health risks, yet little is known about how ordinary users verify such content in real time. In this study, we investigate how behavioral fact-checking strategies, specifically, lateral reading (consulting multiple external sources) versus vertical reading (spending more time on fewer sources), affect accuracy in detecting health-related misinformation. We conducted a large-scale online experiment with 1,842 participants using a designed for the experiment purposes standalone web platform that logs respondents' behavioral signals (tab/window switches and time spent outside the experimental interface). This design enables indirect measurement of fact-checking in an ecologically valid setting. Using linear mixed-effects models with semiparametric bootstrap confidence intervals, we find that lateral reading significantly improves accuracy (β = 0.09, 95% CI [0.079, 0.181], p = 0.001), while vertical reading has no effect (β = -0.01, p = 0.332). Crucially, confirmation bias, operationalized as alignment between statement valence and user attitude, does not moderate this relationship (p = 0.876). This suggests that lateral reading remains effective even when users encounter ideologically congruent misinformation. Our results demonstrate that simple, scalable behavioral cues like tab-switch frequency can serve as reliable indicators of verification quality. For web platforms and digital literacy initiatives, this implies that nudging users to ''open another tab'' may be a lightweight, bias-resistant intervention to improve misinformation resilience, especially in health contexts where errors carry real-world consequences.
Perceived trustworthiness underpins how users navigate online information, yet it remains unclear whether large language models (LLMs), increasingly embedded in search, recommendation, and conversational systems, represent this construct in psychologically coherent ways. We analyze how instruction-tuned LLMs (Llama 3.1 8B, Qwen 2.5 7B, Mistral 7B) encode perceived trustworthiness in web-like narratives using the PEACE-Reviews dataset annotated for cognitive appraisals, emotions, and behavioral intentions. Across models, systematic layer- and head-level activation differences distinguish high- from low-trust texts, revealing that trust cues are implicitly encoded during pretraining. Probing analyses show linearly decodable trust signals and fine-tuning effects that refine rather than restructure these representations. Strongest associations emerge with appraisals of fairness, certainty, and accountability-self– dimensions central to human trust formation online. These findings suggest that modern LLMs internalize psychologically grounded trust signals without explicit supervision, offering a representational foundation for designing credible, transparent, and trustworthy AI systems in the web ecosystem. Code and appendix are available at: https://github.com/GerardYeo/TrustworthinessLLM.
Generative images have proliferated on Web platforms in social media and online copyright distribution scenarios, and semantic watermarking has increasingly been integrated into diffusion models to support reliable provenance tracking and forgery prevention for web content. Traditional noise-layer-based watermarking, however, remains vulnerable to inversion attacks that can recover embedded signals. To mitigate this, recent content-aware semantic watermarking schemes bind watermark signals to high-level image semantics, constraining local edits that would otherwise disrupt global coherence. Yet, large language models (LLMs) possess structured reasoning capabilities that enable targeted exploration of semantic spaces, allowing locally fine-grained but globally coherent semantic alterations that invalidate such bindings. To expose this overlooked vulnerability, we introduce a Coherence-Preserving Semantic Injection (CSI) attack that leverages LLM-guided semantic manipulation under embedding-space similarity constraints. This alignment enforces visual-semantic consistency while selectively perturbing watermark-relevant semantics, ultimately inducing detector misclassification. Extensive empirical results show that CSI consistently outperforms prevailing attack baselines against content-aware semantic watermarking, revealing a fundamental security weakness of current semantic watermark designs when confronted with LLM-driven semantic perturbations.
Most existing content issue detection models are built upon Multimodal Large Language Models (MLLMs). Although such approaches achieve high accuracy, they often suffer from poor timeliness, limited feature richness, and high serving costs: MLLM-based models are typically updated only monthly or quarterly and rely mainly on intrinsic video features (e.g., sampled frames and captions), underutilizing post-hoc user feedback and aggregated video and author-side signals. To address these challenges, we propose a daily updated ranking model for Content Issue Detection based on a Cross-Order Representation Fusion (CORF) architecture. The model supports day-level updates and integrates user behavior features, video- and author-side information, and MLLM-generated scores as inputs. Compared with MLLM-based approaches, our model is significantly smaller and can efficiently score all videos on a daily basis. Experiments on the Adult-Content category show higher recall at the same level of precision, and online A/B tests further demonstrate reduced user exposure to such content issues, contributing to a safer and more positive viewing experience.
With the rise of modern search and recommendation platforms, insufficient collaborative information of cold-start items exacerbates the Matthew effect of existing platform items, challenging platform diversity and becoming a longstanding issue. Existing methods align items' side content with collaborative information to transfer collaborative signals from high-popularity items to cold-start items. However, these methods fail to account for the asymmetry between collaboration and content, nor the fine-grained differences among items. To address these issues, we propose COINS, an item representation enhancement approach based on fused alignment of semantic IDs. Specifically, we use RQ-OPQ encoding to quantize item content and collaborative information, followed by a two-step alignment: RQ encoding transfers shared collaborative signals across items, while OPQ encoding learns items' differentiated information. Comprehensive offline experiments on large-scale industrial datasets demonstrate COINS's superiority, and rigorous online A/B tests confirm statistically significant improvements.
Optimizing numerical systems and mechanism design is crucial for enhancing player experience in Massively Multiplayer Online (MMO) games. Traditional optimization approaches rely on iterative online experiments or parameter tuning over abstracted statistical models, which can be inaccurate, time-consuming and potentially impair players' experience. Although simplified offline simulation systems are frequently employed as alternatives, their low fidelity constrains agents' ability to faithfully replicate real players' reasoning processes and behavioral responses to interventions. To address these limitations, we propose a generative agent-based MMO simulation system with hundreds of agents empowered by Large Language Models (LLMs). By applying Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) on large-scale real player behavioral data, we adapt LLMs from general priors to game-specific domains, enabling realistic and interpretable player decision-making. In parallel, a data-driven environment model trained on real gameplay logs reconstructs dynamic in-game systems. Experiments demonstrate strong consistency with real-world player behaviors and plausible causal responses under interventions, providing a reliable, interpretable, and cost-efficient framework for data-driven numerical design optimization.
We study the auto-bidding problem under a strict return-on-spend constraint (ROSC), where an online algorithm decides how much to bid for each ad slot based on a revealed value and hidden allocation and payment functions. The goal is to maximize cumulative expected utility (value times winning probability) while ensuring that total expected payment does not exceed total expected utility. We prove an impossibility result showing that no online algorithm can achieve sublinear regret even when values, allocations, and payments are drawn i.i.d. from an unknown distribution. For the special case of constant valuations, we design an algorithm that strictly satisfies the ROSC and achieves regret optimal up to logarithmic factors.