Leveraging Large Language Models for Complementary Product Ads Recommendation

Large language models (LLMs) show promise in emergency medicine; however, their role in emergency neurology remains unclear. We developed a customized LLM, Xuanwu-NeuroAid, and prospectively enrolled 433 patients. The diagnostic outputs of the model and emergency physicians were compared with confirmed diagnoses, and an expert panel performed blinded evaluations of the recommendations generated by both the model and physicians. The independent diagnostic accuracy of the model was 79.4%, which was significantly higher than that of emergency physicians (65.4%, p < 0.001). Blinded expert assessments indicated that the model's recommendations for examinations and treatments were significantly more comprehensive, accurate, and clinically applicable than those of physicians (p < 0.001). Moreover, the inclusion of demographic information altered the model's recommendations for health education, suggesting a sensitivity to sociodemographic factors. Our findings highlight the potential of the LLM to enhance diagnostic precision and support decision-making in emergency neurology under simulated clinical conditions. This study was registered at ClinicalTrials.gov (NCT06779292; January 6, 2025).

Recommendation models typically follow a discriminative paradigm, predicting whether items should be retrieved. While effective, the expressive capabilities of these recommender systems are limited. Users can only passively browse the recommended items rather than actively express their needs and engage in an interactive experience. With recent advances in generative models such as large language models, a paradigm shift is happening in the study of recommender systems. Researchers propose building generative recommendation models either by aligning pre-trained generative models with user behaviors or designing recommendation models within a generative framework. These models enable the systems to receive and deliver more human-like content, such as natural language, images, and beyond. In this tutorial, we first provide an overview of the latest progress in generative recommendation models, covering approaches based on large language models, semantic IDs, diffusion models, and more. We then make an in-depth discussion on the challenges, open questions, and potential future directions in developing generative recommendation models.

Recent advancements in Large Language Models (LLMs) have shown significant potential in enhancing recommender systems. However, addressing the cold-start recommendation problem, where users lack historical data, remains a considerable challenge. In this paper, we introduce KALM4Rec (Keyword-driven Retrieval-Augmented Large Language Models for Cold-start User Recommendations), a novel framework specifically designed to tackle this problem by requiring only a few input keywords from users in a practical scenario of cold-start user recommendations. KALM4Rec operates in two main stages: candidates retrieval and LLM-based candidates re-ranking. In the first stage, keyword-driven retrieval models are used to identify potential candidates, addressing LLMs' limitations in processing extensive tokens and reducing the risk of generating misleading information. In the second stage, we employ LLMs with various prompting strategies, including zero-shot and few-shot techniques, to re-rank these candidates by integrating multiple examples directly into the LLM prompts. Our extensive evaluation, using Yelp restaurant data from three English-speaking cities and TripAdvisor hotel data, demonstrates that KALM4Rec excels in improving recommendation quality across the two domains and also highlights its potential for widespread applications.

The rapid growth of biomedical literature poses challenges for manual knowledge curation and synthesis. Biomedical Natural Language Processing (BioNLP) automates the process. While Large Language Models (LLMs) have shown promise in general domains, their effectiveness in BioNLP tasks remains unclear due to limited benchmarks and practical guidelines. We perform a systematic evaluation of four LLMs—GPT and LLaMA representatives—on 12 BioNLP benchmarks across six applications. We compare their zero-shot, few-shot, and fine-tuning performance with the traditional fine-tuning of BERT or BART models. We examine inconsistencies, missing information, hallucinations, and perform cost analysis. Here, we show that traditional fine-tuning outperforms zero- or few-shot LLMs in most tasks. However, closed-source LLMs like GPT-4 excel in reasoning-related tasks such as medical question answering. Open-source LLMs still require fine-tuning to close performance gaps. We find issues like missing information and hallucinations in LLM outputs. These results offer practical insights for applying LLMs in BioNLP.

Large Language Models (LLMs) have achieved significant successes in various research domains by learning the relationship between words. However, while these models are capable of making predictions and inferences based on the learned patterns, they lack logical reasoning abilities, which are crucial for solving problems in both theoretical and practical domains. In addition, traditional logic inference methods are effective in solving problems that are based on logic, but not suitable for general tasks such as recommendations. In response to these challenges, this paper introduces a Logical Large Language Model (L3M) that integrates the strengths of logical reasoning and large language models. The data in L3M is represented in logical expressions and the model uses logical constraints to learn the rules of basic logical operations such as And, Or, and Not. We conduct experiments on both theoretical tasks (solving logical equations) and practical tasks (recommender systems). The results of our theoretical experiments demonstrate that L3M is highly effective in solving logical expressions and variables. Additionally, L3M outperforms the state-of-the-art recommendation models in sequential recommendation tasks.

Various Web API Recommendation (AR) techniques have assisted developers in efficiently identifying suitable APIs for mashup creation. With the emergence of large language models (LLMs), there has been increasing interest in leveraging LLMs for recommender systems. Although several approaches have attempted to utilize LLMs by framing recommendations as prompts, this approach is not ideally suited for AR due to fundamental differences in the training processes of LLMs and AR models. Consequently, it's crucial to conduct further research to identify effective applications of LLMs in AR. To this end, we propose a novel <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LLM</u>-based generative solution for <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">A</u>PI <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</u>ecommendation (LLMAR) that combines instruction learning of multitask and multistage Low-Rank Adaptation fine-tuning based on LLaMA models. Experimental results on the ProgrammableWeb dataset show that LLMAR significantly outperforms representative methods in regular and data-limited scenarios.

Recent advances in Large Language Models (LLMs) have been changing the paradigm of Recommender Systems (RS). However, when items in the recommendation scenarios contain rich textual information, such as product descriptions in online shopping or news headlines on social media, LLMs require longer texts to comprehensively depict the historical user behavior sequence. This poses significant challenges to LLM-based recommenders, such as over-length limitations, extensive time and space overheads, and suboptimal model performance. To this end, in this paper, we design a novel framework for harnessing Large Language Models for Text-Rich Sequential Recommendation (LLM-TRSR). Specifically, we first propose to segment the user historical behaviors and subsequently employ an LLM-based summarizer for summarizing these user behavior blocks. Particularly, drawing inspiration from the successful application of Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) models in user modeling, we introduce two unique summarization techniques in this paper, respectively hierarchical summarization and recurrent summarization. Then, we construct a prompt text encompassing the user preference summary, recent user interactions, and candidate item information into an LLM-based recommender, which is subsequently fine-tuned using Supervised Fine-Tuning (SFT) techniques to yield our final recommendation model. We also use Low-Rank Adaptation (LoRA) for Parameter-Efficient Fine-Tuning (PEFT). We conduct experiments on two public datasets, and the results clearly demonstrate the effectiveness of our approach.

BackgroundHepatocellular carcinoma (HCC) management requires complex decision-making considering tumor burden, liver function, and patient’s functional performance status. Large language models (LLMs) show promise in clinical applications, but their utility in HCC treatment recommendations remains unexplored. We evaluated the clinical relevance of LLM-generated treatment recommendations by comparing concordance with real-world physician decisions and survival outcomes.Methods and findingsWe analyzed 13,614 treatment-naive HCC patients diagnosed between 2008 and 2020 in the Korean Primary Liver Cancer Registry. Treatment recommendations were generated using ChatGPT 4o, Gemini 2.0, and Claude 3.5 with standardized prompts referencing the American Association for the Study of Liver Diseases and the European Association for the Study of the Liver guidelines. Patients were classified as “matched” when LLM recommendations corresponded to actual treatments received. Overall survival (OS) was compared between matched and mismatched groups, stratified by the Barcelona Clinic Liver Cancer (BCLC) stage. Decision tree analysis identified factors influencing treatment selection patterns. Concordance rates between LLM recommendations and physician decisions were 31.1% (ChatGPT 4o), 32.7% (Gemini 2.0), and 26.8% (Claude 3.5). In BCLC-A patients, treatment concordance with LLM recommendations was associated with significantly improved survival (ChatGPT 4o HR: 0.743, 95% CI [0.665, 0.831], P < 0.001). Conversely, in BCLC-C patients, concordance was associated with worse survival outcomes (ChatGPT 4o HR: 1.650, 95% CI [1.523, 1.787], P < 0.001; Gemini 2.0 HR: 1.586, 95% CI [1.470, 1.711], P < 0.001; Claude 3.5 HR 1.483, 95% CI [1.366, 1.610], P < 0.001). In BCLC-B, concordance showed only modest or nonsignificant associations with survival across models. Decision tree analysis revealed that physicians prioritized liver function parameters, while LLMs emphasized tumor characteristics. In early-stage HCC, physicians avoided curative treatments when hepatic reserve was limited, whereas in advanced-stage HCC, physicians preferred locoregional therapies in patients with preserved liver function despite guideline recommendations for systemic therapy. This study is limited by its retrospective design, reliance on registry data without imaging information, and focus on guideline-era treatments, warranting future prospective validation.ConclusionsConcordance between LLM-generated and physician treatment decisions was associated with improved survival in early-stage HCC, whereas this association was not observed in advanced-stage disease. While LLMs may serve as adjunctive tools for guideline-concordant decisions in straightforward scenarios, their recommendations may reflect limited contextual awareness in complex clinical situations requiring individualized care. LLM recommendations should be interpreted cautiously alongside clinical judgment.

BackgroundThe field of natural language processing (NLP) has evolved significantly since its inception in the 1950s, with large language models (LLMs) now playing a crucial role in addressing medical challenges.ObjectivesThis study evaluates the alignment of three prominent LLMs—Gemini, Copilot, and ChatGPT-4.0—with expert consensus on imaging recommendations for acute flank pain.MethodsA total of 29 clinical vignettes representing different combinations of age, sex, pregnancy status, likelihood of stone disease, and alternative diagnoses were posed to the three LLMs (Gemini, Copilot, and ChatGPT-4.0) between March and April 2024. Responses were compared to the consensus recommendations of a multispecialty panel. The primary outcome was the rate of LLM responses matching the majority consensus. Secondary outcomes included alignment with consensus-rated perfect (9/9) or excellent (8/9) responses and agreement with any of the nine panel members.ResultsGemini aligned with the majority consensus in 65.5% of cases, compared to 41.4% for both Copilot and ChatGPT-4.0. In scenarios rated as perfect or excellent by the consensus, Gemini showed 69.5% agreement, significantly higher than Copilot and ChatGPT-4.0, both at 43.4% (p = 0.045 and < 0.001, respectively). Overall, Gemini demonstrated an agreement rate of 82.7% with any of the nine reviewers, indicating superior capability in addressing complex medical inquiries.ConclusionGemini consistently outperformed Copilot and ChatGPT-4.0 in aligning with expert consensus, suggesting its potential as a reliable tool in clinical decision-making. Further research is needed to enhance the reliability and accuracy of LLMs and to address the ethical and legal challenges associated with their integration into healthcare systems.

Large language models (LLMs) are entering clinical workflows, yet their behavior in routine bedside decisions that lack a single "correct" recommendation remains unclear. To describe variation within and across commercially available LLMs when confronted with common, judgment-dependentinpatient medicine management scenarios. Cross-sectional simulation study. Four brief vignettes requiring a binary management decision were posed to each model in five independent sessions. Six LLMs were queried: five general-purpose (GPT-4o, GPT-o1, Claude 3.7 Sonnet, Grok 3, and Gemini 2.0 Flash) and one domain-specific (OpenEvidence). Standardized prompts describing (1) transfusion at borderline hemoglobin, (2) resumption of anticoagulation after gastrointestinal bleed, (3) discharge readiness despite a modest creatinine rise, and (4) peri-procedural bridging in a high-risk patient on apixaban. Primary outcomes were each model's overall recommendation (majority across five runs) and its internal consistency (proportion of identical recommendations across runs; range 0-1). Inter-model agreement was the proportion of models giving the same recommendation. A total of 120 model-vignette interactions were analyzed. Inter-model recommendations diverged in every scenario: transfuse vs observe (67% vs 33% of models), restart vs hold anticoagulation (50% vs 50%), discharge vs delay (50% vs 50%), and bridge vs no-bridge (17% vs 83%). Across five repeated queries of the same vignette, some models changed recommendations in two of five runs (internal consistency as low as 0.60). OpenEvidence was the most internally consistent and concrete in its recommendations; every other model displayed internal variability in one or more vignettes. For nuanced inpatient management questions, widely used LLMs produced inter- and intra-model variation in their recommendations. Clinicians should view LLM output as one perspective among many, consider sampling multiple models or re-prompting, and retain final responsibility for bedside decisions. Prospective studies are needed to test designs that surface model uncertainty and support safe integration of generative AI into complex decision-making.

The prevalence of unhealthy eating habits has become a growing concern in the United States. However, popular food recommendation platforms, such as Yelp, tend to prioritize users' dietary preferences over the healthiness of their choices. While some efforts have focused on developing health-aware food recommendation systems, personalization based on specific health conditions remains underexplored. Additionally, the lack of interpretability in these systems prevents users from evaluating the reliability of recommendations, limiting their practical adoption. To address these issues, we introduce two large-scale personalized health-aware food recommendation benchmarks at the first attempt. Building on this, we propose a novel framework called the Multi-Objective Personalized Interpretable Health-aware Food Recommendation System (MOPI-HFRS). This system generates food recommendations by jointly optimizing three objectives: user preference, personalized healthiness, and nutritional diversity. It also incorporates a reasoning module enhanced by large language models (LLMs) to provide interpretable recommendations that promote healthy dietary knowledge. The framework integrates descriptive features and health data using two structure learning and pooling modules within a graph learning framework. Pareto optimization is applied to balance the multi-faceted objectives. To further enhance healthy dietary knowledge, the system leverages LLMs by infusing knowledge from the recommendation model, generating meaningful interpretations for the recommendations. Extensive experiments on the proposed benchmarks demonstrate that MOPI-HFRS outperforms state-of-the-art methods by delivering diverse, healthy food recommendations alongside reliable explanations.

Food is an essential component in living beings and for that we need agriculture and soil type detection is an essential part of agriculture. To help with that we have created a model where the model gives a hybrid decision support system accurately modeled to reduce the increasing risk of crop failure caused by unstable climatic conditions and soil nutrient differences. When the global agricultural environment drops the dependency in traditional farming, this study shows a strong computational methodology that combines mainly three particular layers to provide highly precise guidance. The starting layer in our model uses Random Forest machine learning model that focuses on soil classification based on soil type. The model then checks essential soil parameters that include Nitrogen (N), Phosphorus (P), and Potassium (K) levels along with the pH. The second layer brings the use of a Geospatial API needed for a 10 year longitudinal climate check in the second layer. The data of historical temperature, rainfall with local soil chemistry helps to calculate the environmental problems. The third layer is the Mistral 7B Large Language Model LLM, which is used in practical farming advice. Machine learning handles quantitative data while the LLM gives a practical context so that it offers agricultural related advice to support data results with real farming experience. The output lists the top five probable crops according to climate changes and difficulty risk scores. This dual focus on soil health and climate gives assurance on productive advice and continuity under volatile situations. Validation results indicate that the model performs almost flawlessly with benchmark data. By connecting the fixed set of probabilistic distribution of the Random Forest with the adaptive intelligence of a Large Language Model, the model provides a comprehensive decision-support tool with fewer limitations from the single-model machine learning applications.

The goal of intent-aware session recommendation (ISR) approaches is to capture user intents within a session for accurate next-item prediction. However, the capability of these approaches is limited by assuming all sessions have a uniform and fixed number of intents. In reality, user sessions can vary, where the number of intentions may differ from one to another. Moreover, they can only learn user intents in the latent space, which further restricts the model's transparency. To ease these issues, we propose a simple yet effective paradigm for ISR motivated by the advanced reasoning capability of large language models (LLMs). Specifically, we first create an initial prompt to instruct LLMs to predict the next item by inferring varying user intents reflected in a session. Then, we propose an effective optimization mechanism to automatically optimize prompts with an iterative self-reflection. Finally, we leverage the robust generalizability of LLMs across diverse domains to efficiently select the optimal prompt for ISR. As such, the proposed paradigm effectively guides LLMs to identify varying user intents at a semantic level, thus delivering more accurate and comprehensible recommendations. Extensive experiments on three real-world datasets verify the superiority of our proposed method.

We present a proof of concept application for interactive recommendations and explanations leveraging the capabilities of Large Language Models (LLMs). The application creates a highly interactive user-driven setting for recommendations giving users the possibility to explicitly tailor recommendations to their needs. Using the possibilities brought by LLMs, the application further generates convincing explanations of recommendations, aligned with the explicitly stated goals of the users. The web application continuously improves by incorporating user feedback and updating recommendations and explanations as needed.

Purpose This study aims to explore the integration of large language models (LLMs) and vision-language models (VLMs) in robotics, highlighting their potential benefits and the safety challenges they introduce, including robustness issues, adversarial vulnerabilities, privacy concerns and ethical implications. Design/methodology/approach This survey conducts a comprehensive analysis of the safety risks associated with LLM- and VLM-powered robotic systems. The authors review existing literature, analyze key challenges, evaluate current mitigation strategies and propose future research directions. Findings The study identifies that ensuring the safety of LLM-/VLM-driven robots requires a multi-faceted approach. While current mitigation strategies address certain risks, gaps remain in real-time monitoring, adversarial robustness and ethical safeguards. Originality/value This study offers a structured and comprehensive overview of the safety challenges in LLM-/VLM-driven robotics. It contributes to ongoing discussions by integrating technical, ethical and regulatory perspectives to guide future advancements in safe and responsible artificial intelligence-driven robotics.