Search Results (10,531)

Generating multiple text sequences and refining them through feedback is essential for improving the quality of outputs in many NLP tasks. While Large Language Models can leverage iterative feedback during inference, smaller models often lack this capability due to limited capacity and the absence of suitable training paradigms. In this paper, we propose a novel Feedback-Aware Inference approach that enables iterative sequence generation with integration of feedback signals. Our method allows models to generate multiple sequences, incorporate feedback from previous iterations, and refine outputs accordingly. This approach dynamically adjusts to different quality metrics, making it adaptable to various contexts and objectives. We evaluate our approach on two distinct tasks: Answer Selection for Question Generation and Keyword Generation, arguing for its generalizability and effectiveness. Results show that our method outperforms strong baselines, maintaining high performance across iterations and achieving superior results even with smaller, open-source models. Full article

Belonging to the Euphorbiaceae family, the Jatropha genus is a promising source for the discovery of antitumor compounds. Jatropha ribifolia is a traditionally used species in folk medicine in the semi-arid region of Brazil, with a few chemical and pharmacological reports. Based on that, the aim of the current work is to isolate, structurally characterize, and assess the cytotoxic activity of isolated compounds through in vitro and in silico analyses. To achieve these main goals, the underground parts were dried, extracted and purified using classical and instrumental chromatographic techniques, leading to the isolation of 16 compounds. Altogether with HR-ESI-MS, IR, one- and two-dimensional NMR experiments, eight previously unreported diterpenes, named ribifolones A-H, along with eight known compounds, were obtained and are herein described. Regarding their activity against melanoma (SK-MEL-28) and colorectal cancer (HCT-116) cell lines, jatrophone was the most potent with IC₅₀ values of 6.19 µM and 10.09 µM, followed by ribifolone C that exhibited a moderate cytotoxicity with IC₅₀ values of 50.71 µM and 33.39 µM, respectively. Network pharmacology analysis suggests the involvement of the PI3K-AKT-mTOR pathway in the activity of both compounds; meanwhile, molecular docking and dynamics simulations demonstrate the main interactions with key proteins in the pathway, indicating putative targets. This work opens new perspectives for the discovery of bioactive compounds found in Euphorbiaceae species, especially from those occurring in Caatinga. Full article

Fish-derived products are extensively acknowledged for their substantial role in fostering balanced diets and supporting a healthy way of life. This research is aimed at formulating, analyzing and evaluating a fish-based food product. The methodology adopted in this study adheres to contemporary food safety standards, prioritizing the utilization of minimal technological processes and natural ingredients, a focus that is gaining prominence within contemporary industrial practices. Thus, the proposal for a formulation obtained by integrating powders and extracts from plant byproducts (Citrus) represents a concrete application direction with real potential for commercialization. The product has been enriched with biocomponents derived from orange peel, namely orange extract (OE) and orange peel powder (PPO). The research focused on product development and the in situ evaluation of the effects of OE and PPO. The physicochemical composition, bioactive compound content, and antioxidant activity were evaluated, along with the microbiological status under post-opening refrigeration conditions, in order to simulate actual consumer use. In addition, the product’s color parameters and sensory attributes were analyzed. The results highlight significant potential for the development of a clean-label fish-based product, characterized by a simplified and easily implementable formulation, aligned with current production and consumption requirements. Compared to the control sample, both OE and PPO significantly influenced the analyzed parameters. Differences in physicochemical composition were observed in the experimental samples. In addition, PPO increased the antioxidant activity of the samples and the profile of bioactive compounds. Microbiological analysis, performed on day 0 and after 3 and 7 days of storage at 4 °C showed opening, confirmed the absence of Escherichia coli and Staphylococcus aureus in all samples and had an influence on the growth of fungi. The acceptability of fish-based products is often limited by odor perception, which is one of the main factors leading to consumer rejection. Sensory evaluation demonstrated that citrus-enriched samples were distinguished by the perception of particular sensory attributes. This formulation presents a practical solution to address this constraint, thereby enhancing the product’s sensory acceptability. The integration of OE and PPO yielded a more harmonized sensory profile, as evidenced by elevated hedonic scores and an intermediate placement in both principal component analysis (PCA) and external preference mapping. This research furnishes a thorough characterization of a fish-based food product, underscoring its potential as a viable option for balanced dietary regimens. Simultaneously, the findings support the product’s adherence to sustainability principles through the utilization of bioactive compounds sourced from plant byproducts, thus satisfying contemporary requirements for foods that possess an optimal nutritional profile and a diminished environmental footprint. Full article

The spread of unverified health claims about drugs, dietary supplements, and alternative remedies on social media poses a growing public health concern. In this study, we present a retrieval-augmented generation (RAG) pipeline that uses large language models (LLMs) grounded in biomedical evidence from PubMed, openFDA adverse event reports, and NIH/NCCIH dietary supplement fact sheets to detect and classify health product misinformation. A total of 3493 health-related posts were collected from Reddit (948 posts across 12 subreddits) and YouTube (2545 video descriptions and comments), from which 8250 structured claims were extracted using Claude Haiku. Each claim was matched to biomedical evidence from three authoritative sources, achieving 79.4% evidence coverage, and classified into one of five veracity categories: supported (7.0%), unsupported (59.9%), exaggerated (22.4%), contradicted (2.0%), or dangerous (8.6%), together with an associated risk tier. Overall, 13.5% of claims were assigned high or critical risk. Cross-platform analysis showed that YouTube contained higher proportions of dangerous (11.3% vs. 2.9%) and exaggerated (27.0% vs. 12.4%) claims than Reddit. Compared with keyword-based and zero-shot transformer baselines, the LLM+RAG pipeline produced a more balanced and fine-grained classification of unsupported, exaggerated, contradicted, and dangerous claims. The most frequently implicated products were ashwagandha, kratom, black seed oil, turmeric, and ivermectin, with disease cure claims showing the highest dangerous classification rate (30.1%). These model-assigned results suggest that evidence-grounded LLM pipelines can support health misinformation surveillance, while also highlighting the need for expert validation and broader cross-platform evaluation. Full article

Against the backdrop of intensifying climate change and deepening digital governance, public service systems in large cities face increasingly severe and complex challenges. Based on multi-source heterogeneous data from large cities in China, this study empirically examines the relationship between climate risk and public service provision and its underlying mechanisms using two-way fixed effects models, mediation models, and threshold regression models. Findings indicate that, first, both physical and transition climate risks are significantly and negatively associated with public service provision. Second, the mediation analysis suggests that physical climate risk is linked to public service provision mainly through entrepreneurial vitality, whereas transition climate risk is linked to public service provision through knowledge spillovers and industrial upgrading. Third, this negative association is more pronounced in coastal cities and cities with larger population scales. Finally, open public data and AI-related development are associated with a partial attenuation of these negative relationships. Therefore, urban policymakers should closely monitor multiple climate-risk pathways, strengthen climate-adaptive governance with data resources and AI technologies, and incorporate differentiated climate vulnerabilities into land-use zoning, green infrastructure planning, and the spatial distribution of critical public services so as to enhance urban resilience. Full article

Modern engineering systems may require multidisciplinary design optimization (MDO) to account for the interactions between coupled physical phenomena. When uncertainties affect model parameters or design variables, these analyses must be extended to uncertainty-based MDO (UMDO), in which objectives and/or constraints are expressed as statistical quantities. However, solving UMDO problems is computationally demanding, especially when costly simulators are involved and the budget must be allocated among uncertainty quantification, multidisciplinary coupling resolution, and optimization. This article introduces a generic multi-fidelity strategy, implemented in the open-source GEMSEO framework, to efficiently address UMDO problems. A fidelity level is defined by a number of samples to estimate the statistics; the higher the fidelity, the higher the number. The strategy solves the UMDO problem for each level by using the solution of the previous level as an initial guess. Numerical experiments are deployed on a simplified overall aircraft design (OAD) problem and a hybrid electric regional aircraft (HERA) case. The results show that, with two fidelity levels, restricting samples and iterations at the low-fidelity stage improves overall performance. This allows the multi-fidelity framework to significantly reduce computational cost compared with single-fidelity approaches (up to 45% for OAD and 40% for HERA) while maintaining or improving solution accuracy. Full article

Millions of persons worldwide experience varying degrees of hearing loss, traditionally addressed through prosthetic solutions such as hearing aids and cochlear implants. However, a significant proportion of individuals cannot benefit from these technologies, cannot access them, or choose not to use them. In this context, non-prosthetic assistive technologies have emerged as a complementary paradigm, leveraging advances in sensing, artificial intelligence, and wearable computing to transform acoustic information into alternative perceptual representations rather than restoring auditory function. This survey provides a review of such systems, focusing on technologies that enhance environmental awareness, communication, and social interaction. Existing approaches are categorized along two main dimensions: the tasks they perform and the platforms on which they operate. Task-oriented analysis includes sound recognition (speech and non-speech), sound source localization, emotion recognition, sign language recognition, and related emerging functionalities. Platform-based analysis emphasizes wearable devices and mobile solutions enabling real-time and context-aware assistance. The survey further highlights key research trends, including real-time auditory scene analysis, portable processing, and artificial intelligence. It shows that recent studies increasingly demonstrate that combining auditory, visual, and haptic modalities improves robustness and usability in real-world conditions, particularly in noisy and dynamic environments. Finally, open challenges such as energy efficiency, latency, evaluation methodologies, and user acceptance are discussed. By synthesizing existing work and identifying open research directions, this survey aims to provide a structured foundation for future developments in intelligent, non-prosthetic assistive systems that redefine how auditory information is accessed and interpreted. Full article

With the rapid development of the Internet, open-source graph datasets are increasingly shared and reused in intelligent networked services, making robust infringement verification increasingly important. Backdoor-based watermarking for graph neural networks (GNNs) can be used to check whether a suspicious model has been trained on protected data without authorization. However, existing dataset infringement verification methods have limited applicability and are mainly designed for private datasets. Directly applying them to open-source datasets would cause models trained by legitimate users to learn backdoor behavior, which would expose them to security risks. In this paper, we propose a new infringement verification method for open-source graph datasets, which reduces backdoor-related security risks in models trained by legitimate users. The core idea is to introduce an additional expansion-class and re-label watermarked samples as belonging to this class. This design completely separates the learning of watermark patterns from the original feature-label mappings during training. As a result, only trigger-bearing samples are directly involved in infringement verification, which helps prevent watermark patterns from being associated with existing classes in the original task. The proposed method provides a practical solution for trustworthy graph data sharing and infringement verification in Internet environments. Extensive experiments on benchmark datasets demonstrate that the proposed method achieves a high verification success rate while largely preserving the model’s clean accuracy. Full article

The knowledge base for many small vertebrate species remains limited, largely because traditional manual data collection methods often overlook less charismatic species, such as reptiles. To address this, our pilot study harnesses open-source deep learning and markerless pose estimation technologies to evaluate the technical feasibility of tracking the spatial use and activity profiles of captive ectotherms. Specifically, we tracked these patterns over two months in a dynamically modified environment for Australian barking geckos (Underwoodisaurus milii). Our findings reveal descriptive changes in spatial occupancy and proximity across varying structural layouts. The system achieved a high raw detection accuracy (96.4%) and spatial categorization accuracy (91.7%) when validated against manual ground-truth data, confirming its robust technical performance and precision. Additionally, we automatically evaluated spatial proxies such as activity time budget, velocity, acceleration, and height usage, standardizing the analysis of extensive video recordings for nocturnal species. This pilot test introduces a simple, cost-effective method for rapid data extraction, offering a reliable, scalable monitoring solution for the management of understudied species. Full article

The intensification of urbanization and the environmental crisis highlight the need for new paradigms of sustainable urban development, in which mobility plays a central role. This article analyzes sustainable urban mobility initiatives in Latin American knowledge territories through a comparative framework that integrates Knowledge-Based Urban Development (KBUD) and urban circularity principles. Grounded in the Fourth-Generation Knowledge Territories (TC4) perspective, the study focuses on governance models and social participation as drivers of transformative mobility practices. Methodologically, it adopts a qualitative and exploratory case study approach, combining primary data from field visits with secondary sources such as legislation, institutional documents, and technical reports. Despite the proliferation of science parks and innovation districts in Latin America, little is known about how governance and social participation shape sustainable mobility initiatives in these contexts, particularly when analyzed through the combined lenses of KBUD and urban circularity. The comparative analysis reveals varying degrees of openness and limitations in urban mobility governance across the three territories selected (distritotec—Mexico, Parque Patricios—Argentina, and Porto Digital—Brazil). The findings reveal distinct governance configurations and degrees of alignment with circular mobility principles. Distritotec stands out for its multistakeholder governance and community-led mobility initiatives, reflecting efforts to operationalize the quintuple helix model. Parque Patricios shows fragmented integration between infrastructure improvements and participatory planning, while Porto Digital presents limited articulation between innovation policies and sustainable mobility, with centralized governance and low public engagement. Persistent challenges observed throughout the cases include the weak institutionalization of citizen participation, insufficient strategies to disincentivize private car use, and a lack of data governance mechanisms. Full article

Empathetic dialogue systems built upon large language models overwhelmingly adopt a monolithic inference paradigm that processes emotion perception, causal reasoning, memory retrieval, and response planning within a single forward pass without architecturally enforced intermediate representations, forfeiting intermediate-state transparency and long-horizon personalization. Drawing on neuroscientific and cognitive–psychological evidence that human empathy is functionally dissociable, we present MOSAIC (Multi-agent Orchestration with Structured Affective memory for Interpretable empathiC dialogue), a training-free framework that operationalizes empathetic dialogue as a four-stage cognitive pipeline: affective perception, causal appraisal, episodic memory retrieval, and response synthesis. Three innovations distinguish MOSAIC from prior work: (1) a cognitively motivated modular architecture whose functionally dissociable stages enable post hoc failure attribution through logged intermediate states; (2) a hierarchical three-tier emotional memory—perceptual, semantic, and episodic—coupled with adaptive three-dimensional retrieval over emotion, situation, and coping-strategy cues; and (3) a heterogeneous model orchestration strategy coordinating open-source and API-accessible models through role-specific chain-of-thought prompts, requiring no task-specific fine-tuning. We note that the EmpatheticDialogues evaluation pre-populates the memory store with 200 training-split episodes prior to test-set interaction, a data-access asymmetry relative to single-model baselines that must be borne in mind when interpreting comparative results. Experiments on EmpatheticDialogues and ESConv show that MOSAIC achieves a 76.4% weighted F1 and an empathy score of 3.87 (on a 1–5 Likert scale) and that it improves over single-model, training-free baselines on aggregate empathy and—most prominently—on human-rated personalization (3.67 vs. 3.24 against Claude-3.5 five-shot, $d=0.48$). We caution that the comparison against training-free baselines is not data access-controlled (see the cold-start discussion in Methods); the personalization advantage, supported by the ablation without the Event Agent, is the result we treat as the primary practical contribution of this work. Full article

Background: Evidence supports the effectiveness and safety of open-source automated insulin delivery (AID) in patients with type 1 diabetes. However, evidence regarding the clinical application of open-source AID in perioperative patients with type 2 diabetes remains limited. Methods: This was an open-label, single-center, exploratory pilot randomized controlled trial (RCT) with parallel groups. Patients with diabetes (excluding type 1 diabetes mellitus) scheduled for elective surgery were randomly assigned to the closed-loop group (open-source hybrid closed-loop AID system) or the control group (conventional insulin pump). The primary outcome was the percentage of time in the target glucose range (TIR, 3.9–10.0 mmol/L). Other efficacy and safety outcomes were also compared between the groups. Results: A total of 49 participants were included and randomized to the closed-loop group (n = 25) or the control group (n = 24). Participants underwent abdominal, orthopedic, thoracic surgery, or neurosurgery during hospitalization. Patients in the closed-loop group had significantly higher TIR than patients in the control group (76.4 ± 14.1% vs. 61.2 ± 20.0%, p = 0.005). Compared with the control group, the closed-loop group also exhibited a 15.6 percentage point reduction in time above range (TAR, >10 mmol/L) without increasing time below range (TBR, <3.9 mmol/L). There were no episodes of severe hypoglycemia (<2.2 mmol/L) or diabetic ketoacidosis in either group. Conclusions: This study demonstrates that in patients with diabetes undergoing elective surgery, the open-source hybrid closed-loop AID system provides better glycemic control than conventional insulin pump therapy. Full article

The incorporation of Distributed Energy Resources (DERs), mainly photovoltaic (PV) systems, creates new challenges for distribution networks, even though these technologies provide significant benefits for decarbonization and grid flexibility. This paper evaluates the impact of high PV penetration on the low-voltage distribution network. The impact was tested on an IEEE 123-bus test network in 24 h simulations. Simulations to evaluate the impacts were conducted using the Open-Source Distribution System Simulator (OpenDSS) and MATLAB via the Component Object Model (COM) interface. The maximum hosting capacity of the different buses was evaluated and then enhanced using smart inverters (SI). The results obtained show improved hosting capacity using fixed lagging PF and Volt-Watt settings. The Volt-Var yielded the worst PV hosting capacity (HC). Full article

Small language models (SLMs) are valued for their computational efficiency and suitability for edge deployment, but often underperform in localized linguistic and cultural contexts due to their limited parameter size. This study explores integrating real-time web search into Qolda, a 4B-parameter Kazakh-centric SLM, to close the performance gap with larger models. We assess two strategies: Naïve Retrieval-Augmented Generation (RAG), which uses raw benchmark questions as search queries, and Query-Refined RAG, which applies various refiner models, including a supervised distillation-tuned Qolda, to optimize queries. On the KazCulture and KazMMLU benchmarks, the Naïve RAG approach in reasoning-enabled mode achieved an average accuracy of 76.00%, improving on the Zero-Shot evaluation result of 60.37%, and, in this system-level comparison, exceeding the Zero-Shot accuracy of larger open-source models such as Qwen3-32B (64.72%) and Gemma-3-27b-it (60.24%), which were evaluated without retrieval augmentation. Query refinement improved the accuracy by about 3%, from 76.00% to 79.46%, but nearly doubled the computational cost. Inference time analysis shows that Naïve RAG adds approximately 1 s of retrieval latency per question. Query refiners introduce up to 4 s of additional overhead. However, the retrieved context reduces the time required for model reasoning in think mode. The most notable gains were observed in localized cultural knowledge, where web search integration correctly answered 32.9% of KazCulture questions that the Zero-Shot baseline failed on, while losing only 16.9% in return. These results suggest that retrieval-augmented SLMs can offer a cost-effective and competitive alternative to larger models for tasks in the domains of Kazakh language and Kazakh culture. Full article

Encryption protects data in the cloud but adds energy cost, especially for partially homomorphic encryption (PHE) schemes that allow computation on encrypted data. Their carbon footprint across cloud data center deployments remains underexplored. We benchmark eight PHE algorithms from the LightPHE open-source Python library, including RSA, ElGamal, Exponential ElGamal, Paillier, Damgård–Jurik, Okamoto–Uchiyama, Goldwasser–Micali, and Elliptic Curve ElGamal, across six cloud environments, and use timing data as input to a carbon estimation model covering Scope 1, Scope 2, and Scope 3 emissions across ten data center configurations. We ground the energy model with a dedicated Intel RAPL calibration on bare-metal hardware using 30 repetitions per configuration. The calibration measures average CPU package power at 34.7 W and total system power at 48.4 W, showing that a fixed 150 W CPU-only assumption overestimates actual CPU power by a factor of 4.3. We present calibrated estimates alongside a 150 W server-class scenario and a sensitivity analysis across power, PUE, and grid carbon intensity. Elliptic curve schemes provide equivalent classical security at a fraction of the energy cost of RSA, and algorithm-specific mathematical structure drives order-of-magnitude differences in carbon output. These results reveal an asymmetry between security and carbon cost across PHE algorithms and establish a sustainable-cryptography baseline for future PQC-based homomorphic schemes. Full article