Despite rapid advances in multimodal large language models, agricultural applications remain constrained by the lack of multilingual speech data, unified multimodal architectures, and comprehensive evaluation benchmarks. To address these challenges, we present AgriGPT-Omni, an agricultural omni-framework that integrates speech, vision, and text in a unified framework.(1) First, we construct a scalable data synthesis and collection pipeline that converts agricultural texts and images into training data, resulting in the largest agricultural speech dataset to date, including 492K synthetic and 1.4K real speech samples across six languages.(2) Second, based on this, we train the first agricultural Omni-model via a three-stage paradigm: textual knowledge injection, progressive multimodal alignment, and GRPO-based reinforcement learning, enabling unified reasoning across languages and modalities.(3) We further propose AgriBench-Omni-2K, the first tri-modal benchmark for agriculture, covering diverse speech–vision–text tasks and multilingual slices, with standardized protocols and reproducible tools. Experiments show that AgriGPT-Omni significantly outperforms general-purpose baselines on multilingual and multimodal reasoning as well as real-world speech understanding.
论文检索
输入标题、作者或关键词，从 17,376 篇学术成果中精准定位
Cultural context profoundly shapes how people interpret online content, yet vision–language models (VLMs) remain predominantly trained through Western or English-centric lenses. This limits their fairness and cross-cultural robustness in tasks like hateful meme detection. We introduce a systematic evaluation framework designed to diagnose and quantify the cross-cultural robustness of state-of-the-art VLMs across multilingual meme datasets, analyzing three axes: (i) learning strategy (zero-shot vs. one-shot), (ii) prompting language (native vs. English), and (iii) translation effects on meaning and detection. Results show that the common ''translate-then-detect'' approach deteriorate performance, while culturally aligned interventions — native-language prompting and one-shot learning — significantly enhance detection. Our findings reveal systematic convergence toward Western safety norms and provide actionable strategies to mitigate such bias, guiding the design of globally robust multimodal moderation systems.
A lesson plan (LP) is a structured guide outlining instructional objectives, methods, and assessments to ensure organized learning. However, existing LP creations are often time-consuming, inconsistent in structure, and lack pedagogical mechanisms for real-time adaptation to diverse learner needs. To address these issues, we propose a co llaborative multi-role a gent framework called COMA for automatic LP generation. COMA formulates LP generation as a collaborative workflow among multiple LLM agents with distinct pedagogical expertise: (1) the novice agent that represents a novice teacher possesses an overarching understanding of the intended lesson flow but demonstrates limited precision in implementing the specific instructional actions; (2) the veteran agent that represents an experienced teacher demonstrates deep familiarity with the curriculum, textbooks, and the knowledge components embedded in each unit; and (3) the master agent that represents a pedagogical expert exhibits a well-developed and confident grasp of lesson progression, with the ability to design, adapt, and implement specific instructional actions effectively and responsively. Through an iterative workflow, these agents collaboratively refine LP quality. Comprehensive experiments across five subjects, using expert-designed metrics, demonstrate that COMA significantly outperforms state-of-the-art methods, producing lesson plans with superior quality, coherence, and pedagogical alignment. Our framework offers a robust solution for generating deployable instructional content at scale. Data and code are available at https://github.com/ai4ed/COMA-LessonPlan.
Large Language Models (LLMs) are now widely deployed across modern web services, but their safe and trustworthy use in real-world settings critically depends on accurate alignment with human preferences. Preference alignment is typically achieved using methods such as reinforcement learning or direct preference optimization (DPO), whose effectiveness in practice hinges on the quality of labeled preference data. However, a fundamental practical challenge remains: preference datasets inevitably contain noise. Through a systematic analysis of mainstream preference datasets, we find that roughly 25% of preference pairs show clear inconsistencies between reward-model evaluations and human annotations. Such inconsistent examples do not convey reliable preference signals; training directly on them therefore not only fails to improve alignment but can even degrade model behavior. To address this problem, we propose Noise-Aware Preference Alignment for LLMs via Confidence and Polarity Reweighting (AlignCP), a fully automated, human-free framework for noise-aware preference alignment. AlignCP derives two interpretable metrics from reward-model outputs: Confidence, which measures the reliability of each preference judgment, and Polarity, which evaluates whether the reward-model ranking agrees with the original human label. These metrics are combined to assign a training weight to each sample—amplifying high-confidence, label-consistent pairs while down-weighting or discarding low-confidence, contradictory, or otherwise noisy pairs. Unlike methods that rely on human re-inspection, repeated relabeling, or heavy computational reruns, AlignCP performs automated data-quality control with minimal overhead and no human intervention. Empirical results show that AlignCP substantially outperforms existing data-centric alignment approaches on standard preference benchmarks and remains more robust under noisy supervision.
As large language models (LLMs) increasingly shape content generation, interaction, and decision-making across the Web, aligning them with human values has become a central objective in trustworthy AI. This challenge becomes even more pronounced when aligning multiple, potentially conflicting human values. Although recent approaches, such as reward reweighting, prompt-based supervised fine-tuning, and model merging, attempt to tackle multi-value alignment, they still face two major limitations: (1) training separate models for each value combination is prohibitively expensive; (2) value conflicts substantially degrade alignment performance. These limitations make it difficult to achieve favorable trade-offs across diverse human values. To address these challenges, we revisit multi-value alignment from the perspective of value consistency in data and propose VC-soup, a data filtering and parameter merging framework grounded in value-consistent learning. We first design a value consistency metric based on the cosine similarity between the reward-gap vector of each preference pair and an all-ones vector, which quantifies its cross-value coherence. We then filter out low-consistency preference pairs in each value dataset and train on the remaining data to obtain smooth, value-consistent policy models that better preserve linear mode connectivity. Finally, we linearly combine these policies and apply Pareto filtering across values to obtain solutions with balanced multi-value performance. Extensive experiments and theoretical analysis demonstrate that VC-soup effectively mitigates conflicts and consistently outperforms existing multi-value alignment methods.
Access to mental healthcare is increasingly strained by workforce shortages and rising demand, motivating the development of intelligent systems that can support mental healthcare experts. We introduce coTherapist, a unified framework utilizing a small language model to emulate core therapeutic competencies through domain-specific fine-tuning, retrieval augmentation, and agentic reasoning. Evaluation on clinical queries demonstrates that coTherapist generates more relevant and clinically grounded responses than contemporary baselines. Using our novel T-BARS rubric and psychometric profiling, we confirm coTherapist exhibits high empathy and therapist-consistent personality traits. Furthermore, human evaluation by domain experts validates that coTherapist delivers accurate, trustworthy, and safe responses. coTherapist was deployed and tested by clinical experts. Collectively, these findings demonstrate that small models can be engineered to exhibit expert-like behavior, offering a scalable pathway for digital mental health tools.
Misinformation on the web increasingly appears in multimodal forms, combining text, images, and OCR-rendered content in ways that amplify harm to public trust and vulnerable communities. While prior fact-checking systems often rely on unimodal signals or shallow fusion strategies, modern misinformation campaigns operate across modalities and require models that can reason over subtle cross-modal inconsistencies in a transparent and responsible manner. We introduce MultiCheck, a lightweight and interpretable framework for multimodal fact verification that jointly analyzes textual, visual, and OCR evidence. At its core, MultiCheck employs a relational fusion module based on element-wise difference and product operations, allowing for explicit cross-modal interaction modeling with minimal computational overhead. A contrastive alignment objective further helps the model distinguish between supporting and refuting evidence while maintaining a small memory and energy footprint, making it suitable for low-resource deployment. Evaluated on the Factify-2 (5-class) and MOCHEG (3-class) benchmarks, MultiCheck achieves substantial performance improvement and remains robust under noisy OCR and missing modality conditions. Overall, MultiCheck is efficient, interpretable, and robust for multimodal verification. Our code is available at the following: https://github.com/Adityakishore09/MultiCheck_WWW-2026 GitHub repository.
Child mental health screening faces growing challenges from rising psychological problems and limited professional access. Traditional self-report tools rely on verbal ability and self-awareness, limiting their validity in younger children. Projective drawing tests offer a nonverbal alternative, with the Draw-A-Person (DAP) test widely used to elicit psychological cues from drawings. While translating DAP test into Web-based screening, multimodal large language models (MLLMs) are emerging as a key enabling mechanism. However, their ability to deliver construct-level clarity and interpretive consistency for responsible deployment has not been systematically evaluated. To address this gap, we propose DAPWeb, a construct-aligned evaluation framework that assesses whether MLLMs can reliably support DAP-based child mental screening in Web environments. DAPWeb introduces a clinically grounded benchmark derived from real drawings and defines task-structured evaluation across six psychological constructs and three essential screening abilities: determination, detection, and comparison. The metrics emphasize construct validity and cross-drawing consistency, reflecting real-world early-screening workflows. Experiments across MLLMs reveal substantial gaps from human experts in most abilities. Yet, the comparable performance on determination suggests that DAP test can serve as a feasible component of Web-based early screening under structured interpretation. Thus, DAPWeb provides a replicable paradigm for responsible Web AI in child mental health.
Web-based platforms are becoming a primary channel for psychological support, yet most LLM-driven chatbots remain opaque, single-stage, and weakly grounded in established therapeutic practice. To address this gap, we present XInsight, a multi-agent framework that models psychological support as a stage-consistent workflow aligned with the classical Exploration-Insight-Action paradigm. Building on structured client representations, XInsight orchestrates specialized agents under a unified Reason-Intervene-Reflect cycle: an Exploration agent organizes background and concerns into a structured Case Conceptualization Form, a Routing agent performs Adaptive Therapeutic Routing (ATR) across SFBT, CBT, and MBCT, a unified Therapeutic agent executes school-consistent submodules, and a Consolidation agent guides review, skill integration, and relapse-prevention planning. A Recording agent continuously transforms open-ended web dialogues into standardized psychological artifacts, enhancing interpretability, continuity, and accountability. To support transparent assessment, we introduce XInsight-Bench with a Scale-Guided LLM Evaluation (SGLE) protocol that combines therapy-specific clinical scales with general counseling criteria. Experiments show improved paradigm alignment, multi-therapy integration, interaction depth, and interpretability over existing multi-agent counseling systems, indicating that XInsight provides a practical blueprint for integrating counseling-inspired support agents into web applications for digital well-being.
Precision pathology relies on detecting fine-grained morphological abnormalities within specific Regions of Interest (ROIs), as these local, texture-rich cues—rather than global slide contexts—drive expert diagnostic reasoning. While Vision-Language (V-L) models promise data efficiency by leveraging semantic priors, adapting them faces a critical Granularity Mismatch, where generic representations fail to resolve such subtle defects. Current adaptation methods often treat modalities as independent streams, failing to ground semantic prompts in ROI-specific visual contexts. To bridge this gap, we propose the Hierarchical Adaptation and Alignment Framework (HAAF ). At its core is a novel Cross-Level Scaled Alignment (CLSA) mechanism that enforces a sequential calibration order: visual features first inject context into text prompts to generate content-adaptive descriptors, which then spatially guide the visual encoder to spotlight anomalies. Additionally, a dual-branch inference strategy integrates semantic scores with geometric prototypes to ensure stability in few-shot settings. Experiments on four benchmarks show HAAF significantly outperforms state-of-the-art methods and effectively scales with domain-specific backbones (e.g., CONCH) in low-resource scenarios.
As web platforms evolve towards greater personalization and emotional complexity, conversational agents must transcend superficial empathy to demonstrate identity-aware emotional reasoning. However, existing systems face two limitations: (1) reliance on situation-centric datasets lacking persistent user identity, which hampers the capture of personalized affective nuances; and (2) dependence on opaque, coarse reward signals that hinder development of verifiable empathetic reasoning. To address these gaps, we introduce KardiaBench, a large-scale user-grounded benchmark comprising 178,080 QA pairs across 22,080 multi-turn conversations anchored to 671 real-world profiles. The dataset is constructed via a model-in-the-loop pipeline with iterative rubric-guided refinement to ensure psychological plausibility and persona consistency. This progressive empathy pipeline that integrates user comprehension, contextual reasoning, and emotion perception into conversations, followed by iterative critique and rubric-based refinement to ensure psychological plausibility, emotional fidelity, and persona consistency. Building on this, we propose Kardia-R1, a framework that trains models for interpretable, stepwise empathetic cognition. Kardia-R1 leverages Rubric-as-Judge Empathetic Reinforcement Learning (Rubric-ERL), a GRPO-based method that uses explainable, human-aligned rubric rewards to tightly couple user understanding, emotional inference, and supportive response generation. Extensive experiments across four LLM backbones demonstrate that Kardia-R1 consistently outperforms other methods in emotion accuracy, empathy, relevance, persona consistency, and safety.
The vision of an inclusive World Wide Web is impeded by a severe linguistic divide, particularly for communities in low-resource regions of Southeast Asia. While large language models (LLMs) offer a potential solution for translation, their deployment in data-poor contexts faces a dual challenge: the scarcity of high-quality, culturally relevant data and the prohibitive energy costs of training on massive, noisy web corpora. To resolve the tension between digital inclusion and environmental sustainability, we introduce Sustainable Agent-Guided Expert-tuning (SAGE). This framework pioneers an energy-aware paradigm that prioritizes the ''right data'' over ''big data''. Instead of carbon-intensive training on unfiltered datasets, SAGE employs a reinforcement learning (RL) agent, optimized via Group Relative Policy Optimization (GRPO), to autonomously curate a compact training set. The agent utilizes a semantic reward signal derived from a small, expert-constructed set of community dialogues to filter out noise and cultural misalignment. We then efficiently fine-tune open-source LLMs on this curated data using Low-Rank Adaptation (LoRA). We applied SAGE to translation tasks between English and seven low-resource languages (LRLs) in Southeast Asia. Our approach establishes new state-of-the-art performance on BLEU-4 and COMET-22 metrics, effectively capturing local linguistic nuances. Crucially, SAGE surpasses baselines trained on full datasets while reducing data usage by 97.1% and training energy consumption by 95.2%. By delivering high-performance models with a minimal environmental footprint, SAGE offers a scalable and responsible pathway to bridge the digital divide in the Global South.
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.
Dengue, a mosquito-borne disease, continues to pose a persistent public health challenge in urban areas, particularly in tropical regions such as Singapore. Effective and affordable control requires anticipating where transmission risks are likely to emerge so that interventions can be deployed proactively rather than reactively. This study introduces a novel framework that uncovers and exploits latent transmission links between urban regions, mined directly from publicly available dengue case data. Instead of treating cases as isolated reports, we model how hotspot formation in one area is influenced by epidemic dynamics in neighboring regions. While mosquito movement is highly localized, long-distance transmission is often driven by human mobility, and in our case study, the learned network aligns closely with commuting flows, providing an interpretable explanation for citywide spread. These hidden links are optimized through gradient descent and used not only to forecast hotspot status but also to verify the consistency of spreading patterns, by examining the stability of the inferred network across consecutive weeks. Case studies on Singapore during 2013–2018 and 2020 show that four weeks of hotspot history are sufficient to achieve an average F-score of 0.79. Even under the COVID-19 ''circuit breaker,'' when mobility patterns were severely disrupted, the model remained robust with an F-score of 0.83. Importantly, the learned transmission links align with commuting flows, highlighting the interpretable interplay between hidden epidemic spread and human mobility. By shifting from simply reporting dengue cases to mining and validating hidden spreading dynamics, this work transforms open web-based case data into a predictive and explanatory resource. The proposed framework advances epidemic modeling while providing a scalable, low-cost tool for public health planning, early intervention, and urban resilience.
Live streaming platforms require real-time monitoring and reaction to social signals, utilizing partial and asynchronous evidence from video, text, and audio. We propose StreamSense, a streaming detector that couples a lightweight streaming encoder with selective routing to a Vision–Language Model (VLM) expert. StreamSense handles most timestamps with the lightweight streaming encoder, escalates hard/ambiguous cases to the VLM, and defers decisions when context is insufficient. The encoder is trained using (i) a cross-modal contrastive term to align visual/audio cues with textual signals, and (ii) an IoU-weighted loss that down-weights poorly overlapping target segments, mitigating label interference across segment boundaries. We evaluate StreamSense on multiple social streaming detection tasks (e.g., sentiment classification and hate content moderation), and the results show that StreamSense achieves higher accuracy than VLM-only streaming while only occasionally invoking the VLM, thereby reducing average latency and compute. Our results indicate that selective escalation and deferral are effective primitives for understanding streaming social tasks. Code is publicly available on GitHub.
Accurate rainfall forecasting is essential for climate and disaster management, but precipitation exhibits extreme zero inflation that modern time-series Foundation Models (TSFMs) fundamentally cannot represent due to their continuous regression outputs. This structural mismatch causes pervasive drizzle-like false alarms, miscalibrated nonzero intensities, and severely underdetected extremes, while retraining large TSFMs is computationally prohibitive and environmentally unsustainable for most regions. We present a training-free wrapper that corrects zero inflation for frozen TSFMs without updating any parameters. Our method restores discrete zero mass using empirical occurrence statistics, aligns positive-value distributions via probability-integral transforms, and applies Generalized Pareto tail mapping for extreme-value consistency. Experiments on South Australian rainfall show substantial gains with negligible overhead (<5,ms per forecast, compared to hundreds of GPU-hours for retraining). The proposed wrapper enables carbon-neutral, globally deployable climate services and directly advances the goals of UN SDG~13 (Climate Action).
For web platforms facing regulatory scrutiny---from content moderation to ad delivery and recommendations---fairness audits routinely disagree due to metric choice, subgroup granularity, sampling variance, and dataset shift. Point estimates yield brittle pass/fail narratives that are hard to defend in governance contexts. We propose a Bayesian audit-of-audits that pools heterogeneous audits---count-based and metric-only---into interval-valued fairness claims with explicit uncertainty and policy-risk tables aligned to practitioner thresholds. The framework unifies classification and exposure metrics, enforces consistency across coarse and intersectional group definitions via soft coherence constraints, and quantifies the Value-of-Information of prospective audits. We also provide heterogeneity diagnostics and leave-one-audit-out sensitivities. Across a synthetic Audit Zoo, a content-moderation case study on CivilComments--WILDS, and an ad-delivery simulation, our meta-evaluator attains near-nominal coverage with narrower intervals and fewer decision flips than per-audit baselines, while integrating partial-information audits.
Table retrieval is the task of retrieving the most relevant tables from large-scale corpora given natural language queries. However, structural and semantic discrepancies between unstructured text and structured tables make embedding alignment particularly challenging. Recent methods such as QGpT attempt to enrich table semantics by generating synthetic queries, yet they still rely on coarse partial-table sampling and simple fusion strategies, which limit semantic diversity and hinder effective query–table alignment. We propose STAR (Semantic Table Representation), a lightweight framework that improves semantic table representation through semantic clustering and weighted fusion. STAR first applies header-aware K-means clustering to group semantically similar rows and selects representative centroid instances to construct a diverse partial table. It then generates cluster-specific synthetic queries to comprehensively cover the table's semantic space. Finally, STAR employs weighted fusion strategies to integrate table and query embeddings, enabling fine-grained semantic alignment. This design enables STAR to capture complementary information from structured and textual sources, improving the expressiveness of table representations. Experiments on five benchmarks show that STAR achieves consistently higher Recall than QGpT on all datasets, demonstrating the effectiveness of semantic clustering and weighted fusion for robust table representation. Our code is available at https://github.com/adsl135789/STAR.
Large Language Models (LLMs) perform well across many tasks but degrade when processing large collections of repetitive or highly similar inputs, a common scenario in applications such as near-duplicate search results and large e-commerce catalogs. In these settings, concatenation-based approaches—long-context prompting and supervised fine-tuning—suffer from attention saturation and diminished signal-to-noise ratio, causing models to miss subtle but important distinctions as input size grows. We introduce SEER (Set Encoding for Efficient Representation), a framework that enables LLMs to handle massive sets of near-duplicate items through a single learned token. SEER first encodes individual items with a pretrained embedding model, then aggregates them using an adapter that captures inter-item relationships and preserves fine-grained differences while mitigating redundancy. To ensure both discriminative and generative capabilities, we propose a multi-task alignment strategy that supervises set-level descriptions across multiple semantic dimensions. Experiments on a large-scale e-commerce dataset demonstrate that SEER substantially outperforms in-context and fine-tuned LLM baselines, maintaining stable performance even when processing thousands of highly similar items. These results establish SEER as an effective and scalable approach for LLM processing of dense, redundant input sets.
Topic modeling plays a critical role in organizing and understanding large-scale web content. While neural topic models (NTMs) based on variational autoencoders (VAEs) have achieved notable success in analyzing textual data, they remain limited in addressing the multimodal nature of modern web content. Existing unimodal or multimodal extensions often suffer from posterior collapse and fail to capture the directional semantics inherent in both text and images, resulting in incoherent topics and limited interpretability. To address these challenges, we propose MM-vNTM (MultiModal Neural Topic Model with von Mises-Fisher Mixtures), a framework for web-scale topic discovery over multimodal data. MM-vNTM leverages pre-aligned cross-modal embeddings as inputs and jointly models document-level representations of text and image modalities in a shared hyperspherical latent space. Furthermore, it defines topics as mixtures of von Mises-Fisher (vMF) distributions in the L2-normalized word embedding space, explicitly capturing directional similarity. Experiments on multimedia web datasets demonstrate that MM-vNTM consistently outperforms state-of-the-art unimodal and multimodal baselines in terms of overall topic quality, highlighting its effectiveness for real-world web scenarios.