Generative search systems are increasingly replacing link-based retrieval with AI-generated summaries, yet little is known about how these systems differ in sources, language, and fidelity to cited material. We examine responses to 11,000 real search queries across five systems—vanilla GPT, Search GPT, Perplexity Search with Grok, Google AI Overviews, and traditional Google Search—at three levels: source diversity, linguistic characterization of the generated summary, and source-summary fidelity. We find that generative search systems exhibit significant source-selection biases in their citations, favoring certain sources over others. Incorporating search also selectively attenuates epistemic markers, reducing hedging by up to 60% while preserving confidence language in the AI-generated summaries. At the same time, AI summaries further compound the citation biases: Wikipedia and longer sources are disproportionately overrepresented, whereas cited social media content and negatively framed sources are substantially underrepresented. Our findings highlight the potential for answer bubbles, in which identical queries yield structurally different information realities across systems, with implications for user trust, source visibility, and the transparency of AI-mediated information access.

Multi-agent LLM systems often coordinate by compressing an upstream interaction into a handoff artifact that downstream agents treat as shared state. We show that this handoff step is a structural source of privacy leakage: summaries preferentially preserve operational facts while weakening the boundary metadata that governs how those facts may be used—a failure mode we call summary collapse. On a controlled multi-agent coordination testbed we measure marker survival with a human-validated judge (𝜅 = 0.74), where 𝜎_b = 1 means every boundary marker survives verbatim and 𝜎_b = 0 means all are lost. Boundary-marker and operational-fact survival are nearly uncorrelated at the handoff level on both GPT-5-mini and DeepSeek-R1-32B (Pearson r near zero): uncompressed free-text handoffs preserve boundaries at 𝜎_b ≈ 0.80, whereas a 25-word budget drops 𝜎_b to ≈0.57 while operational-fact survival stays near ceiling. Controlled downstream tests reveal that protection depends on boundary explicitness: vague languages leak in 73% of GPT and 50% of DeepSeek cases, while explicit constraints reduce leakage to under 15% across all three tested models. A no-handoff single-agent control further shows the failure is not reducible to multi-agent topology as direct full-marker access still leaks more often than the operationalized handoff. Prompt-only mitigation and exact-string redaction only partially address the problem, while a gold-derived audience allowlist nearly eliminates leakage across models, showing that correctly identifying audience boundaries is the key factor.

Large language models (LLMs) frequently generate toxic content, posing significant risks for safe deployment. Current mitigation strategies often degrade generation quality or require costly human annotation. We propose CausalDetox, a framework that identifies and intervenes on the specific attention heads causally responsible for toxic generation. Using the Probability of Necessity and Sufficiency (PNS), we isolate a minimal set of heads that are necessary and sufficient for toxicity. We utilize these components via two complementary strategies: (1) Local Inference-Time Intervention, which constructs dynamic, input-specific steering vectors for context-aware detoxification, and (2) PNS-Guided Fine-Tuning, which permanently unlearns toxic representations. We also introduceParaTox, a novel benchmark of aligned toxic/non-toxic sentence pairs enabling controlled counterfactual evaluation. Experiments on ToxiGen, ImplicitHate, and ParaDetox show that CausalDetox achieves up to 5.34% greater toxicity reduction compared to baselines while preserving linguistic fluency, and offers a 7× speedup in head selection.

Dense retrievers in retrieval-augmented generation (RAG) systems exhibit systematic biases—including brevity, position, literal matching, and repetition biases—that can compromise retrieval quality. Query rewriting techniques are now standard in RAG pipelines, yet their impact on these biases remains unexplored. We present the first systematic study of how query enhancement techniques affect dense retrieval biases, evaluating five methods across six retrievers. Our findings reveal that simple LLM-based rewriting achieves the strongest aggregate bias reduction (54%), yet fails under adversarial conditions where multiple biases combine. Mechanistic analysis uncovers two distinct mechanisms: simple rewriting reduces bias through increased score variance, while pseudo-document methods achieve reduction through genuine decorrelation from bias-inducing features. However, no technique uniformly addresses all biases, and effects vary substantially across retrievers. Our results provide practical guidance for selecting query enhancement strategies based on specific bias vulnerabilities. More broadly, we establish a taxonomy distinguishing query-document interaction biases from document encoding biases, clarifying the limits of query-side interventions for debiasing RAG systems.

News consumption on social media has become ubiquitous, yet how different forms of engagement shape psychosocial outcomes remains unclear. To address this gap, we leveraged a large-scale dataset of ~26M posts and ~45M comments on the Bluesky platform and conducted a quasi-experimental study, matching 81,345 Treated users exposed to News feeds with 83,711 Control users using stratified propensity score analysis. We examined psychosocial wellbeing in terms of affective, behavioral, and cognitive outcomes. Our findings reveal that news engagement produces systematic trade-offs: increased depression, stress, and anxiety, yet decreased loneliness and increased social interaction on the platform. Regression models reveal that News feed bookmarking is associated with greater psychosocial deterioration compared to commenting or quoting, with magnitude differences exceeding tenfold. These per-engagement effects accumulate with repeated exposure, showing significant psychosocial impacts. Our work extends theories of news effects beyond crisis-centric frameworks by demonstrating that routine consumption creates distinct psychological dynamics depending on engagement type, and bears implications for tools and interventions for mitigating the psychosocial costs of news consumption on social media.

Online communities develop distinct norms for content they collectively value, yet it remains unclear whether current language models can recognize locally valued contributions in context. We formalize this as community-conditioned preference prediction and introduce Vastu (Value-Aware Social Tuning), a benchmark of 75,000 Reddit comments from 15 communities spanning Gaming, Science, Q&A, Advice, and Politics. We evaluate four model families—prompted LLMs, LoRA-adapted SLMs, supervised encoders, and feature-based classifiers—across global, local, and context-conditioned settings. Our central finding is that parametric adaptation consistently outperforms prompting: supervised encoders reach 0.74 AUROC and fine-tuned SLMs 0.64–0.71, while the best prompted result is only 0.62. This gap is not merely quantitative—vanilla prompting yields over 80% false-negative rates, systematically discarding content communities actually value. Conversational context narrows but does not close this divide. Together, these results suggest that local preference recognition requires community-specific training signal, not just better prompting. Our work supports future research on community-aware reward modeling, feed curation, and positive moderation.

Large language models (LLMs) have shown promise in many natural language understanding tasks, including content moderation. However, these models can be expensive to query in real-time and do not allow for a community-specific approach to content moderation. To address these challenges, we explore the use of open-source small language models (SLMs) for community-specific content moderation tasks. We fine-tune and evaluate SLMs (less than 15B parameters) by comparing their performance against much larger open- and closed-sourced models in both a zero-shot and few-shot setting. Using 150K comments from 15 popular Reddit communities, we find that SLMs outperform zero-shot LLMs at content moderation-11.5% higher accuracy and 25.7% higher recall on average across all communities. Moreover, few-shot in-context learning leads to only a marginal increase in the performance of LLMs, still lacking compared to SLMs. We further show the promise of cross-community content moderation, which has implications for new communities and the development of cross-platform moderation techniques. Finally, we outline directions for future work on language model based content moderation.

Large language models (LLMs) are now ubiquitous in user-facing applications, yet they still generate undesirable toxic outputs, including profanity, vulgarity, and derogatory remarks. Although numerous detoxification methods exist, most apply broad, surface-level fixes and can therefore easily be circumvented by jailbreak attacks. In this paper we leverage sparse autoencoders (SAEs) to identify toxicity-related directions in the residual stream of models and perform targeted activation steering using the corresponding decoder vectors. We introduce three tiers of steering aggressiveness and evaluate them on GPT-2 Small and Gemma-2-2B, revealing trade-offs between toxicity reduction and language fluency. At stronger steering strengths, these causal interventions surpass competitive baselines in reducing toxicity by up to 20%, though fluency can degrade noticeably on GPT-2 Small depending on the aggressiveness. Crucially, standard NLP benchmark scores upon steering remain stable, indicating that the model’s knowledge and general abilities are preserved. We further show that feature-splitting in wider SAEs hampers safety interventions, underscoring the importance of disentangled feature learning. Our findings highlight both the promise and the current limitations of SAE-based causal interventions for LLM detoxification, further suggesting practical guidelines for safer language-model deployment.

Large language models (LLMs) have shown great potential in flagging harmful content in online communities. Yet, existing approaches for moderation require a separate model for every community and are opaque in their decision-making, limiting real-world adoption. We introduce Mixture of Moderation Experts (MoMoE), a modular, cross-community framework that adds post-hoc explanations to enable scalable content moderation. MoMoE orchestrates four operators—Allocate, Predict, Aggregate, Explain—and is instantiated as seven community-specialized experts (MoMoE-Community) and five norm-violation experts (MoMoE-NormVio). On 30 unseen subreddits, the best variants obtain Micro-F1 scores of 0.72 and 0.67, respectively, matching or surpassing strong fine-tuned baselines while consistently producing concise and reliable explanations. Although community-specialized experts deliver the highest peak accuracy, norm-violation experts provide steadier performance across domains. These findings show that MoMoE yields scalable, transparent moderation without needing per-community fine-tuning. More broadly, they suggest that lightweight, explainable expert ensembles can guide future NLP and HCI research on trustworthy human-AI governance of online communities.

Key point extraction is an important task in argument summarization which involves extracting high-level short summaries from arguments. Existing approaches for KP extraction have been mostly evaluated on the popular ArgKP21 dataset. In this paper, we highlight some of the major limitations of the ArgKP21 dataset and demonstrate the need for new benchmarks that are more representative of actual human conversations. Using SoTA large language models (LLMs), we curate a new argument key point extraction dataset called ArgCMV comprising of ∼12K arguments from actual online human debates spread across ∼3K topics. Our dataset exhibits higher complexity such as longer, co-referencing arguments, higher presence of subjective discourse units, and a larger range of topics over ArgKP21. We show that existing methods do not adapt well to ArgCMV and provide extensive benchmark results by experimenting with existing baselines and latest open source models. This work introduces a novel KP extraction dataset for long-context online discussions, setting the stage for the next generation of LLM-driven summarization research.

Accurately simulating human opinion dynamics is crucial for understanding a variety of societal phenomena, including polarization and the spread of misinformation. However, the agent-based models (ABMs) commonly used for such simulations often over-simplify human behavior. We propose a new approach to simulating opinion dynamics based on populations of Large Language Models (LLMs). Our findings reveal a strong inherent bias in LLM agents towards producing accurate information, leading simulated agents to consensus in line with scientific reality. This bias limits their utility for understanding resistance to consensus views on issues like climate change. After inducing confirmation bias through prompt engineering, however, we observed opinion fragmentation in line with existing agent-based modeling and opinion dynamics research. These insights highlight the promise and limitations of LLM agents in this domain and suggest a path forward: refining LLMs with real-world discourse to better simulate the evolution of human beliefs.

Creating human-like large language model (LLM) agents is crucial for faithful social simulation. Having LLMs role-play based on demographic information sometimes improves human likeness but often does not. This study assessed whether LLM alignment with human behavior can be improved by integrating information from empirically-derived human belief networks. Using data from a human survey, we estimated a belief network encompassing 64 topics loading on nine non-overlapping latent factors. We then seeded LLM-based agents with an opinion on one topic, and assessed the alignment of its expressed opinions on remaining test topics with corresponding human data. Role-playing based on demographic information alone did not align LLM and human opinions, but seeding the agent with a single belief greatly improved alignment for topics related in the belief network, and not for topics outside the network. These results suggest a novel path for human-LLM belief alignment in work seeking to simulate and understand patterns of belief distributions in society.