Integrating LLM Usage in Gamified Systems

Despite the impressive performance across numerous tasks, Large Language Models (LLMs) often fail in solving simple decision-making tasks due to the misalignment of the knowledge in LLMs with environments. On the contrary, Reinforcement Learning (RL) agents learn policies from scratch, which makes them always align with environments but difficult to incorporate prior knowledge for efficient explorations. To narrow the gap, we propose TWOSOME, a novel general online framework that deploys LLMs as decision-making agents to efficiently interact and align with embodied environments via RL without requiring any prepared datasets or prior knowledge of the environments. First, we query the joint probabilities of each valid action with LLMs to form behavior policies. Then, to enhance the stability and robustness of the policies, we propose two normalization methods and summarize four prompt design principles. Finally, we design a novel parameter-efficient training architecture where the actor and critic share one frozen LLM equipped with LOw-Rank Adapters (LoRA) updated by PPO. We conduct extensive experiments to evaluate TWOSOME. (i) TWOSOME exhibits significantly better sample efficiency and performance compared to the conventional RL method, PPO, and prompt tuning method, SayCan, in both classical decision-making environment, Overcooked, and simulated household environment, VirtualHome. (ii) Benefiting from LLMs’ open-vocabulary feature, TWOSOME shows superior generalization ability to unseen tasks. (iii) Under our framework, there is no significant loss of the LLMs’ original ability during online PPO finetuning.a

Background: Virtual coaching can help people adopt new healthful behaviors by encouraging them to set specific goals and helping them review their progress. One challenge in creating such systems is analyzing clients’ statements about their activities. Limiting people to selecting among predefined answers detracts from the naturalness of conversations and user engagement. Large Language Models (LLMs) offer the promise of covering a wide range of expressions. However, using an LLM for simple entity extraction would not necessarily perform better than functions coded in a programming language, while creating higher long-term costs. Methods: This study uses a real data set of annotated human coaching dialogs to develop LLM-based models for two training scenarios: one that generates pattern-matching functions and the other which does direct extraction. We use models of different sizes and complexity, including Meta-Llama, Gemma, and ChatGPT, and calculate their speed and accuracy. Results: LLM-generated pattern-matching functions took an average of 10 milliseconds (ms) per item as compared to 900 ms. (ChatGPT 3.5 Turbo) to 5 s (Llama 2 70B). The accuracy for pattern matching was 99% on real data, while LLM accuracy ranged from 90% (Llama 2 70B) to 100% (ChatGPT 3.5 Turbo), on both real and synthetically generated examples created for fine-tuning. Conclusions: These findings suggest promising directions for future research that combines both methods (reserving the LLM for cases that cannot be matched directly) or that use LLMs to generate synthetic training data with more expressive variety which can be used to improve the coverage of either generated codes or fine-tuned models.

Burnout has become a significant concern in high-stress industries, impacting mental health and productivity. Traditional burnout management solutions lack real-time adaptability and personalized intervention strategies. This research explores an AI-driven burnout management mechanism leveraging Generative AI, MongoDB Atlas, Python, and Large Language Models (LLMs) to provide real-time detection, personalized coping strategies, and predictive analytics. The system collects and stores user responses, behavioral patterns, and sentiment data in MongoDB Atlas, while LLMs analyze and generate tailored burnout mitigation recommendations. Experimental results demonstrate the effectiveness of the proposed approach in early burnout detection, personalized intervention, and user engagement, showing improvements in stress reduction and work-life balance. The proof of concept for AI-powered burnout detection and intervention demonstrates the technical feasibility of using Large Language Models (LLMs), Generative AI, MongoDB Atlas, and Python-based automation for real-time mental health support. The system successfully classifies burnout severity using LLMs trained on sentiment analysis, achieving an accuracy of 91.2%, which significantly outperforms traditional rule-based sentiment models. The use of MongoDB Atlas for scalable data storage ensures real-time burnout tracking, while FastAPI-based backend automation allows for seamless integration across platforms. Furthermore, Generative AI provides personalized stress interventions, dynamically adapting to user feedback and improving engagement. These technical validations highlight how AI-driven burnout management can be scalable, real-time, and enterprise-ready, paving the way for AI-powered diagnostics in mental health applications across various industries.

Access to specialty surgical care is growing in many low-income countries, but it remains unclear how hospital workforces can leverage technology to manage large numbers of increasingly complex patients. Large language models (LLMs) may be helpful for this type of clinical decision support, but their real-world performance and safety remain uncertain. The objective of this study was to evaluate feasibility, usability, and potential benefits and risks of an LLM-based assistant for postoperative neurosurgical care in the Gambia. A prospective, single-arm implementation study was conducted at the Edward Francis Small Teaching Hospital. A convenience sample of 4 medical officers (MOs) and 5 nurses assigned to the neurosurgical service participated. A prompted GPT-4o Turbo was deployed on OpenAI Pro accounts to support performance. Usability, helpfulness, and safety were the primary outcomes. Cost-effectiveness was a secondary outcome. Participants completed 75 LLM-assisted interactions on 9 postoperative neurosurgery patients. Usability metrics indicated a moderately high cognitive workload, marginally acceptable usability of the LLM system, and high perceived ease of use. Management plan quality improved in 45 of 75 mock rounds interactions (60%), with a mean improvement of 8.5% (P < .001) on mock rounds scoring rubrics. The improvement was greater for MOs (21.0% change) than nurses (6.5% change). In hypothetical case dilemmas, MO plan accuracy improved by 22.7% (P = .001), and critical errors declined from 33.3% to 0%. Fourteen care changes for 9 patients were attributed to LLM suggestions, including 6 that potentially prevented major morbidity. No unsafe outputs were detected. Exploratory cost analysis suggested potential savings from clinical care changes exceeded the labor costs involved in LLM use. LLM use was associated with improved plan quality without observed safety concerns, while also prompting clinically meaningful care changes. Larger, controlled studies are needed to determine generalizability, durability of benefit, and patient-centered outcomes.

The rapid advancement of Large Language Models (LLMs) has opened new possibilities for analyzing Web server access logs, particularly in the e-commerce domain. Web server logs provide a rich source of data for system monitoring, security assessment, and user behavior analysis. LLMs may facilitate automated log parsing, anomaly detection, and user session analysis, reducing manual effort, typically required for this kind of tasks. The paper investigates the applicability of a popular LLM, ChatGPT-4o, for Web server log data analysis. A series of experiments on a real e-commerce server log was performed, focusing on six key areas: general statistics and automated reporting, user session reconstruction and analysis, error diagnosis and anomaly detection, Web bot detection, generating synthetic user sessions, and analysis of e-customer activities. The results demonstrate the efficiency of LLMs in extracting meaningful insights from raw log data without extensive pre-processing, making advanced log analysis more accessible to researchers and practitioners. However, challenges such as necessity of precise prompt formulation, the risk of LLM-generated hallucinations, and generalized interpretations require careful validation. The study provides practical guidelines for integrating LLMs into e-commerce server log analysis, emphasizing their advantages, limitations, and potential business applications.

BackgroundGestational diabetes mellitus (GDM) is a prevalent chronic condition that affects maternal and fetal health outcomes worldwide, increasingly in underserved populations. While generative artificial intelligence (AI) and large language models (LLMs) have shown promise in health care, their application in GDM management remains underexplored.ObjectiveThis study aimed to investigate whether retrieval-augmented generation techniques, when combined with knowledge graphs (KGs), could improve the contextual relevance and accuracy of AI-driven clinical decision support. For this, we developed and validated a graph-based retrieval-augmented generation (GraphRAG)–enabled local LLM as a clinical support tool for GDM management, assessing its performance against open-source LLM tools.MethodsA prototype clinical AI assistant was developed using a GraphRAG constructed from 1212 peer-reviewed research articles on GDM interventions, retrieved from the Semantic Scholar API (2000‐2024). The GraphRAG prototype integrated entity extraction, KG construction using Neo4j, and retrieval-augmented response generation. The performance was evaluated in a simulated environment using clinical and layperson prompts, comparing the outputs of the systems against ChatGPT (OpenAI), Claude (Anthropic), and BioMistral models across 5 common natural language generation metrics.ResultsThe GraphRAG-enabled local LLM showed higher accuracy in generating clinically relevant responses. It achieved a bilingual evaluation understudy score of 0.99, Jaccard similarity of 0.98, and BERTScore of 0.98, outperforming the benchmark LLMs. The prototype also produced accurate, evidence-based recommendations for clinicians and patients, demonstrating its feasibility as a clinical support tool.ConclusionsGraphRAG-enabled local LLMs show much potential for improving personalized GDM care by integrating domain-specific evidence and contextual retrieval. Our prototype proof-of-concept serves two purposes: (1) the local LLM architecture gives practitioners from underserved locations access to state-of-the-art medical research in the treatment of chronic conditions and (2) the KG schema may be feasibly built on peer-reviewed, indexed publications, devoid of hallucinations and contextualized with patient data. We conclude that advanced AI techniques such as KGs, retrieval-augmented generation, and local LLMs improve GDM management decisions and other similar conditions and advance equitable health care delivery in resource-constrained health care environments.

This study investigated the feasibility of using large language models (LLMs) to infer problematic Instagram use, which refers to excessive or compulsive engagement with the platform that negatively impacts users’ daily functioning, productivity, or well-being, from a limited set of metrics of user engagement in the platform. Specifically, we explored whether OpenAI’s GPT-4o and Google’s Gemini 1.5 Pro could accurately predict self-reported problematic use tendencies based solely on readily available user engagement metrics like daily time spent on the platform, weekly posts and stories, and follower/following counts. Our sample comprised 775 Italian Instagram users (61.6% female; aged 18–63), who were recruited through a snowball sampling method. Item-level and total scores derived by querying the LLMs’ application programming interfaces were correlated with self-report items and the total score measured via an adapted Bergen Social Media Addiction Scale. LLM-inferred scores showed positive correlations with both item-level and total scores for problematic Instagram use. The strongest correlations were observed for the total scores, with GPT-4o achieving a correlation of r = 0.414 and Gemini 1.5 Pro achieving a correlation of r = 0.319. In cross-validated regression analyses, adding LLM-generated scores, especially from GPT-4o, significantly improved the prediction of problematic Instagram use compared to using usage metrics alone. GPT-4o’s performance in random forest models was comparable to models trained directly on Instagram metrics, demonstrating its ability to capture complex, non-linear relationships indicative of addiction without needing extensive model training. This study provides compelling preliminary evidence for the use of LLMs in inferring problematic Instagram use from limited data points, opening exciting new avenues for research and intervention.

Recommender systems (RecSys) have become critical tools for enhancing user engagement by delivering personalized content across diverse digital platforms. Recent advancements in large language models (LLMs) demonstrate significant potential for improving RecSys, primarily due to their exceptional generalization capabilities and sophisticated contextual understanding, which facilitate the generation of flexible and interpretable recommendations. However, the direct deployment of LLMs as primary recommendation policies presents notable challenges, including persistent latency issues stemming from frequent API calls and inherent model limitations such as hallucinations and biases. To address these issues, this paper proposes a novel offline reinforcement learning (RL) framework that leverages imitation learning from LLM-generated trajectories. Specifically, inverse reinforcement learning is employed to extract robust reward models from LLM demonstrations. This approach negates the need for LLM fine-tuning, thereby substantially reducing computational overhead. Simultaneously, the RL policy is guided by the cumulative rewards derived from these demonstrations, effectively transferring the semantic insights captured by the LLM. Comprehensive experiments conducted on two benchmark datasets validate the effectiveness of the proposed method, demonstrating superior performance when compared against state-of-the-art RL-based and in-context learning baselines. The code can be found at https://github.com/ArronDZhang/IL-Rec.

Cross-category recommendation systems aim to suggest items to users from various categories, enhancing the user experience by broadening their exposure to diverse products or services. Large Language Models (LLMs) offer a promising avenue for improving the effectiveness of such systems. LLMs, trained on vast datasets, excel at understanding complex relationships between different categories of products by analysing patterns in user interactions, preferences, and item descriptions. By leveraging natural language understanding, these models can identify similarities and connections between disparate product categories, enabling more personalized and relevant cross-category recommendations. This approach goes beyond traditional recommendation algorithms that typically focus on a single category or rely heavily on collaborative filtering techniques. Instead, LLMs can process and integrate information from multiple domains, making it easier to suggest complementary products from different categories that users may not have initially considered. Additionally, LLMs can interpret contextual information, such as user reviews and sentiment analysis, to further refine recommendations based on the user's unique preferences and behavioural patterns. The potential of LLMs in cross-category recommendations opens new possibilities for e-commerce, entertainment, and other industries where personalized experiences are key to user engagement and satisfaction. As these models continue to evolve, they offer the promise of more intelligent and comprehensive recommendation systems that can better meet the needs of users by offering diverse and relevant suggestions across categories. These abstract outlines the key role of LLMs in advancing cross-category recommendation systems, highlighting their capacity to improve personalization and broaden the scope of recommendations.

The RAG Chatbot can be considered an AI document-interactive system that integrates Retrieval-Augmented Generation (RAG) technology, thus allowing the owner to interact with their archival documents. It uses Streamlit for its user interface with the backend employing Ollama for the local running of the Large Language Model (LLM) and FAISS for fast vector-based similarity search. The users can upload their documents in formats such as PDF, DOCX, and TXT which are then directly subjected to the text extraction, chunking, and embedding generation process. The embeddings that are produced are indexed by FAISS which allows the chatbot to retrieve the document parts relevant to its answer. The system architecture assures that the response generated by the LLM is always tied to the user's document improving precision and enabling the customer to identify the information source. Besides, the chatbot has features such as document management in real-time, debug mode for response tracing, and local data persistence that not only augment its functionalities but also guarantee privacy. This project illustrates a complete end-to-end realization of a local RAG pipeline integrating language comprehension, retrieval speed, and user engagement in a seamless conversational interface.

The integration of Large Language Models (LLMs) with simulations represents a novel frontier in computational modeling, offering unprecedented opportunities to improve the realism, user experience, and analytical capabilities of simulated environments. This paper explores the synergistic potential of combining LLMs, such as GPT-4, with various types of simulations to create more dynamic, responsive, and intelligent systems. Beyond enhancing realism, LLMs can improve user engagement, provide advanced analysis of simulation data, act as consultants within interactive simulations, serve as active agents, assist in debriefing and decision-making processes, and even aid in the development of simulations.

BACKGROUND Electronic cigarette (e-cigarette) use is prevalent in youth and young adults. TikTok, a popular social media platform for youth and young adults, has been used to disseminate e-cigarette-related videos, primarily dominated by promotional videos. OBJECTIVE We aim to identify key e-cigarette-related TikTok video features associated with high user engagement to assist with future video design for vaping prevention campaigns. METHODS We collected 1,487 e-cigarette-related TikTok videos and related metadata using the TikTok API (Application Programming Interface). We applied large language models GPT-4 and Video-LLaMA to extract video features (e.g., promotion content, background, gender, lifestyle, talking, cartoon, vaping tricks, containing emoji) from e-cigarette-related TikTok videos. We randomly selected and hand-coded 25 videos to check the accuracy of two models in identifying these video features. We utilized generalized linear models with identity link functions to identify significant video features associated with high TikTok user engagement (likes + shares + comments)/views. RESULTS Compared to the Video-LLaMA model, the GPT-4 model exhibited higher accuracy (83%-100 % vs. 24%-88 %) in video feature identification. Notably, video backgrounds in cars, private spaces, or shops demonstrated significantly higher user engagement than in public spaces. Moreover, videos featuring young adults, smoking or vaping, talking, vape tricks, containing emojis, or funny and silly content exhibited heightened user engagement. Conversely, videos with promotional content or featuring e-cigarettes experienced lower engagement. CONCLUSIONS TikTok video features like background settings, young adult presence, talking, and containing emojis substantially enhance user engagement. These insights offer valuable guidance for designing compelling videos in vaping prevention campaigns to improve social media user engagement.

We developed an innovative system that combines Natural Language Understanding (NLU), a curated knowledge base, and the efficient management of a Large Language Model (LLM) to support motivational health coaching. Using Rasa as the core framework, we enhanced it by integrating the GPT-3.5-turbo model. Users opt into reflective dialogues during conversations. When they respond to open-ended questions, their input goes directly to the GPT-3.5-turbo model, allowing for more flexible responses. To provide curated trustworthy content, we integrated a knowledge provision component that searches a PDF-based knowledge base and generates user-friendly responses using Retrieval-Augmented Generation. We tested the system in a real-world scenario by deploying it on a Nao robot in seven older adults’ homes for 1–2 weeks, encouraging positive behavioral changes in some users. Our system serves as a valuable foundation for building an even more integrated, personalized system that can connect with other Application Programing Interfaces (APIs) and integrate with home sensors and edge devices.

Leveraging the robust potential inherent in large language models, their profound and pervasive impact has transcended various domains in recent years, ushering in their widespread integration across diverse sectors. In resonance with this predominant trend, the current study introduces an innovative application that endows users with the capacity to actively engage in conversations with their digital files. This program integrates state-of-the-art large language models with the techniques of Retrieval Augmentation, thereby crafting an immersive and sophisticated experience that not only amplifies but fundamentally elevates user engagement to new heights of interactivity and responsiveness. Functioning as a pivotal nexus, Hugging Face, a renowned platform for machine learning models, assumes the role of the primary repository and catalyst for these transformative language models. Through the medium of this application, users can have interactive engagement, perfectly aligned with the continually evolving tapestry of linguistic technology and digital interaction. Significantly, users possess the freedom to choose from an extensive array of open-source large language models available on the Hugging Face platform, thereby, they also retain the option to seamlessly update to newer models as they become available, ensuring continuous access to the latest large language models and maintaining the applicability of the application in line with evolving user needs. Importantly, the operational viability of the program is extended to local execution, contingent upon the availability of sufficient hardware resources.

This paper proposes the utilization of large language models as recommendation systems for museum visitors. Since the aforementioned models lack the notion of context, they cannot work with temporal information that is often present in recommendations for cultural environments (e.g., special exhibitions or events). In this respect, the current work aims to enhance the capabilities of large language models through a fine-tuning process that incorporates contextual information and user instructions. The resulting models are expected to be capable of providing personalized recommendations that are aligned with user preferences and desires. More specifically, Generative Pre-trained Transformer 4, a knowledge-based large language model is fine-tuned and turned into a context-aware recommendation system, adapting its suggestions based on user input and specific contextual factors such as location, time of visit, and other relevant parameters. The effectiveness of the proposed approach is evaluated through certain user studies, which ensure an improved user experience and engagement within the museum environment.