Browse papers

461 papers

Understanding International Patient Experiences in South Korean Plastic Surgery: An AI-Assisted Analysis of Patient Reviews, Service Quality, Communication, and Trust

Ishaan Rastogi · Aug 15, 2026

South Korea is one of the most visible destinations in global cosmetic-surgery tourism, and the experience patients report there depends on far more than the surgical result: communication, consultation quality, pricing, staff interaction, and post-operative support all shape how a procedure is remembered and described. This paper reports a mixed-methods study of international patient experience in South Korean plastic surgery, built around a dataset of 74 patient narratives and service ratings across 23 clinics, assembled in three stages: a hand-compiled research document (9 clinics, 36 observations), a Google Maps-sourced expansion (6 clinics, 20 observations), and a deliberately non-Google expansion (8 clinics, 18 observations) drawn from a medical-tourism marketplace (WhatClinic), a Korean-language patient community (Gangnam Unni), a Korean booking/discount platform (Babitalk), a Korean review aggregator (Sungyesa), a medical-tourism referral marketplace (Bookimed), and a public social-media post (Threads). Every clinic in the combined dataset has at least three of five service dimensions populated by real review content. Price remained the most consistently low-scoring dimension (lowest or tied-lowest at 11 of 22 comparable clinics), while staff interaction and post-operative care were the most consistently praised. The three-stage collection design produced a direct, measured finding about data-collection method itself: the Google-sourced batch contained zero mixed-sentiment observations, while the non-Google batch recovered three, suggesting the earlier "no mixed sentiment" pattern was a property of Google's specific review-panel selection rather than of web-sourced data in general. The expansion also surfaced a substantive negative review alleging undisclosed additional charges before anesthesia, a documented historical "ghost doctor" controversy independently corroborated against a 2016 Korea Times report[2], and, via Bookimed[3], an unusually long first-person account of an entire multi-week patient journey that included a frightening but ultimately normal post-operative complication. An AI-assisted lexicon classifier was run against all 74 narratives and checked against human-coded labels, reaching 82% sentiment accuracy and a macro F1 of 0.74 on theme classification; a specific, honestly-reported failure mode emerged where the classifier's keyword-based "review credibility" trigger over-fires on the phrase "verified reviewer," a phrasing artifact of how the newer batches were written up rather than a deeper semantic problem. The paper does not rank clinics or make claims about surgical safety; its contribution is a research-based account of what shapes international patient experience, a demonstration of how data-source diversification changes what a review dataset can say, and an assessment of where AI-assisted analysis helps and where it needs human oversight.

Plastic SurgeryLarge Language ModelsMedical TourismSouth KoreaSentiment Analysis

Technical Feasibility Analysis of a Low-Cost UAV-Based Computer Vision System for Wildlife Monitoring

jeevanr17 · Aug 15, 2026

Abstract. Wildlife monitoring in forested and environmentally challenging regions requires observation systems that can provide sufficient coverage while reducing the dependence on continuous manual surveillance. Unmanned aerial vehicles (UAVs) combined with computer vision provide a potential approach for automated wildlife observation; however, practical deployment is constrained by factors such as payload capacity, propulsion requirements, wireless communication, computational resources, environmental conditions, and detection performance. This paper presents a technical feasibility analysis of a low-cost UAV-based wildlife monitoring system integrating a Pixhawk 2.4.8 flight controller, GPS-assisted navigation, an ESP32-CAM for image acquisition, Wi-Fi-based video transmission, and YOLOv8s for animal detection. The detection model was trained for ten wildlife classes using a dataset containing 5,553 images and 8,014 annotated object instances, with 4,442 images used for training and 1,111 for validation. The model was trained for 100 epochs at an image size of 640 × 640 pixels with a batch size of 8. At the final training epoch, the model achieved a validation precision of 94.01%, recall of 94.01%, mAP@50 of 96.76%, and mAP@50–95 of 83.54%. The system was also examined at the system level with respect to wireless communication, image acquisition, and environmental conditions. The UAV configuration uses A2212/13T 1000-KV brushless motors, 10 × 4.5 propellers, and an 11.1-V 5400-mAh Li-Po battery, with a UAV mass of approximately 2.0 kg and a considered payload range of 1.2–1.5 kg. Engineering calculations are used to examine the propulsion requirements for the resulting take-off mass, while flight endurance is treated as a theoretical parameter because an experimentally measured endurance value was not established. The analysis evaluates the technical feasibility of integrating low-cost UAV hardware with external computer-vision processing while identifying communication dependence, low-light degradation, and the absence of onboard AI processing as important limitations.

YOLOv8RoboticsUAVComputer VisionWildlife Monitoring

SQS: A Multi-Dimensional Framework for Evaluating Software Specifications in AI-Assisted Development

Abhavya Anekverna · Aug 15, 2026

Abstract If AI coding agents need to make a lot of inferences, they are more likely to create incorrect code, but if they have less work to do, theoretically it should be more accurate. In an AI coding agent with little to no human interaction, a specification that doesn't specify a behavior can be interpreted as a decision on implementation, with the agent continuing to work without asking for clarification. If a decision on such a behavior, constraint, interface, or acceptance condition is made, it is called a specification-induced assumption, and it is therefore an implementation decision, distinct from an ordinary implementation decision the specification leaves open (e.g., what front-end framework to use). We hypothesize that the quality of a specification impacts the correctness of implementation in the downstream, as follows: Lower quality specification = fewer assumptions an agent may have to make = less deviation from what the specification actually specifies. Current requirements-quality research has been geared towards human-centred requirements processes, and the studies of ambiguity, completeness and consistency have concentrated on them separately. This paper introduces the Specification Quality Score (SQS), a rubric-anchored, operational measure based on five dimensions of specification quality (Completeness, Unambiguity, Internal Consistency, Intent Traceability, Machine-Actionability) with clear criteria for each band of each dimension. We provide worked examples of the framework and a clear explanation of the meaning of the total score. This is a conceptual paper – it is not an empirical validation study, and does not imply that SQS predicts code generation results. This correlational statement, and the controlled experiment necessary to test it, is explicitly suggested as the next step. Keywords: requirements engineering, specification quality, large language models, AI coding agents, specification-driven development

Requirements EngineeringSoftware EngineeringNatural Language ProcessingArtificial IntelligenceComputer Science

FuelGuard AI:An AI-Driven Unified Framework for Intelligent Fuel and LPG Allocation During Emergencies and Supply Shortages

Akshay N, Akshay Guptha L, Akshatha H M, Aishwarya S · Aug 15, 2026

Abstract. Fuel and LPG are essential resources for transportation, households, healthcare, agriculture, and other critical services. During emergencies, supply shortages, and sudden demand fluctuations, conventional distribution systems may face challenges in forecasting demand, managing limited resources, and ensuring fair and secure access. This paper presents FuelGuard AI, an intelligent digital framework developed to support fuel and LPG allocation through centralized monitoring, quota management, priority-based allocation, demand analysis, and secure transaction verification. The implemented prototype integrates citizen and station-operator interfaces with a secure QR-based authentication mechanism using short-lived cryptographic tokens, enabling authorized operators to verify allocation requests before dispensing fuel. The system also maintains transaction records and supports server-side quota deduction to improve traceability and prevent unauthorized or duplicate allocations. In addition, the framework provides a foundation for AI-assisted demand forecasting, shortage detection, and adaptive allocation during critical situations. The proposed system aims to improve transparency, security, resource utilization, and fairness while complementing existing fuel and LPG distribution infrastructure.

Energy ManagementCybersecurityResource AllocationArtificial IntelligenceFuel Management

UrbanNexus: A Digital Twin-Enabled Framework for Multi-Objective Evaluation of Sustainable Urban Interventions

Yashika Rajput · Aug 15, 2026

Abstract Rapid urbanization has intensified interconnected challenges such as air pollution, urban heat, inadequate green cover, and increasing pressure on urban infrastructure. Although artificial intelligence and geospatial technologies have enabled effective prediction and monitoring of individual urban phenomena, existing approaches often address these problems independently and provide limited support for comparing alternative interventions before their real-world implementation. This creates a need for decision-support systems capable of evaluating environmental, health, economic, and implementation trade-offs within a unified framework. This paper proposes UrbanNexus, a Digital Twin-enabled urban decision intelligence framework for the multi-objective evaluation of sustainable urban interventions. The framework integrates heterogeneous environmental, geospatial, demographic, and infrastructure data with machine learning-based urban-state prediction and a simulation-oriented Digital Twin. Alternative interventions, such as tree plantation, green infrastructure, electric public transportation, and hybrid strategies, are represented as scenarios and evaluated according to their potential environmental benefits, health impacts, implementation costs, feasibility, and deployment requirements. A multi-objective decision layer identifies trade-offs among competing strategies, while an explainable decision layer provides interpretable justification for scenario rankings and recommendations. A supporting agentic layer is proposed to coordinate the interpretation of outputs from environmental, health, economic, optimization, and explanation components. The primary contribution of UrbanNexus is a unified framework for cross-domain, multi-objective evaluation and explainable ranking of heterogeneous urban interventions before deployment. The proposed framework establishes the architecture, mathematical decision formulation, and experimental protocol required for future empirical validation. Keywords: Urban AI, Digital Twin, Decision Intelligence, Sustainable Urban Planning, Multi-Objective Optimization, Scenario Simulation, Explainable AI, Multi-Agent Systems

Digital TwinsMulti-Objective OptimizationSustainable CitiesArtificial IntelligenceUrban Planning

A Mutation-Based Framework for Assessing Code Understanding Through Behavioral Change

Manish Kumar · Aug 15, 2026

A learner who can correctly predict what a piece of code will output has shown one skill; a learner who can also predict how that output changes after a small edit, and explain why, has shown something closer to real understanding. This paper builds an assessment around that second skill. Each item pairs an original program with a small, behavior-changing modification of it, evaluates both on the same fixed input, and then asks four questions in sequence: what the original program outputs, what the modified program outputs, whether the two outputs differ, and why. Conventional mutation testing uses the same kind of small, artificial change but points it at a test suite rather than a person — a mutant is "killed" if some test detects it, and the exercise measures test-suite adequacy. Here the mutant is aimed at the participant instead: it becomes the question, and the participant's response is what gets studied. The resulting instrument has seven items and a maximum score of 35 points, backed by a Python pipeline that generates the assessment, validates incoming responses, scores both the objective and free-text answers, checks the dataset for integrity problems, and produces summary statistics. No human participants have been recruited yet, so the pipeline was instead run against a clearly labeled synthetic pilot — 20 simulated respondent profiles across three ability bands, yielding 140 item-level responses. That run produced a mean score of 25.55/35 (73.0%), band means of 20.43/35, 23.29/35, and 34.17/35 for the Beginner, Intermediate, and Strong profiles, and objective-item accuracies of 80.7%, 74.3%, and 80.0% on the first three questions. None of these numbers describe real learners; they describe whether the pipeline itself behaves correctly, which it did. What this paper contributes, then, is not a set of findings but a validated instrument and pipeline, ready to be pointed at real participants. Keywords: code understanding; program comprehension; mutation analysis; mutation testing; behavioral change; programming assessment; reproducible pipeline; synthetic pilot

Mutation TestingCode Comprehensionsoftware EngineeringProgramming EducationComputer Science

Skin Cancer Detection Using Preprocessed Vision Transformers: An Interpretability-Driven Approach

Ananya Atri · Aug 15, 2026

Abstract : Skin cancer is an issue affecting health all over the world, and proper diagnosis of the issue would lead to better patient outcomes. Deep Learning, in particular Convolutional Neural Networks (CNNs), has brought more objective and fast screening techniques. This paper assesses the diagnostic capability of Vision Transformers (ViT) with Explainable AI (XAI) on HAM10000 data (10,015 pigmented skin lesion images). The researchers used a critically important preprocessing pipeline comprising of morphology transformations to remove hair, image segmentation to identify regions of interest and data augmentation to overcome imbalanced classes. Five XAI techniques were used, namely, Grad-CAM, LIME, SHAP, Integrated Gradients and Saliency Maps to improve diagnostic transparency. The experiment outcomes revealed that ViT + Grad-CAM architecture was the best performer with a total accuracy of 96.6%. The model achieved 0.92 precision, 0.90 recall and 0.92 F1-score. Grad-CAM was especially useful in a clinical setting as it is fast to run and localizes classes accurately.

Skin CancerMedicineComputer VisionArtificial IntelligenceVision Transformer

PreClinAI: A Multi-Parametric Simulation Engine for Dynamic Physiological Trajectories

Nabeela Javed · Aug 15, 2026

Abstract. A purely computational pipeline for preclinical drug evaluation would allow toxicity profiles and experimental protocols to be generated directly from molecular structures without premature reliance on live subjects. Standard in-silico models provide part of the solution, but the main ethical and financial benefits are lost if arbitrary in-vivo testing is still required to determine systemic physiological outcomes. We propose a solution to preclinical trial inefficiency using a multi-parametric digital twin framework. The system evaluates molecular features through a calibrated classifier to predict toxicity probabilities, then maps these risks onto simulated biological trajectories modulated by strain genetics, age, and baseline organ health. The resulting physiological simulation allows for precise statistical power analysis, generating the absolute minimum animal sample size required to achieve significance. To ensure interpretability, the architecture grounds its findings by mapping the predicted endpoints against a deterministic biological knowledge graph, forming a mechanistic explanation backed by literature retrieval. The framework requires minimal preliminary in-vivo data, allowing researchers to systematically reduce and refine animal testing models before physical trials commence.

Systems PharmacologyComputataional BiologyComputaional Toxicology

Signaling Competence in Global Remote Technical Hiring: An Open-Source Framework

Tanvi · Aug 15, 2026

Abstract: The old way of hiring people is not working well when it comes to finding good software engineers who can work from anywhere. This is because the world is changing and more people are working from home.When companies try to hire people from countries it is hard for them to know if the person is really good at their job. This is a problem because companies do not want to hire someone who is not very good. We think that the things people do online like the code they write and the projects they work on can show if they are really good at their job. We also think that how well someone can communicate with others even when they are not in the place is very important. We looked at a lot of data. Found out that the things people do online are a much better way to know if they are good at their job than the school they went to or the companies they used to work for. We also found out that when people work on open-source projects it is a sign that they are very good at their job.We came up with a way for companies to hire people that is based on what they can actually do, not just what they say they can do. This way is better because it is fair and it helps companies find the people for the job.In the end we think that this new way of hiring people will make it easier for companies to find software engineers and will help people get jobs that they arereally good, at. It will also help reduce the costs of hiring people and make sure that everyone has a chance of getting a job.

Signaling Competence in Global Remote Technical Hiring: An Open-Source Framework

Signalling Competence in Global Remote Technical Hiring: An Open-Source Framework

Anushka Pujari · Aug 15, 2026

Abstract: The old way of hiring people is not working well when it comes to finding good software engineers who can work from anywhere. This is because the world is changing and more people are working from home. When companies try to hire people from other countries, it is hard for them to know if the person is really good at their job. This is a problem because companies do not want to hire someone who is not very good. We think that the things people do online, like the code they write and the projects they work on, can show if they are really good at their job. We also think that how well someone can communicate with others, even when they are not in the place, is very important. We looked at a lot of data. Found out that the things people do online are a much better way to know if they are good at their job than the school they went to or the companies they used to work for. We also found out that when people work on open-source projects, it is a sign that they are very good at their job. We came up with a way for companies to hire people that is based on what they can actually do, not just what they say they can do. This way is better because it is fair and it helps companies find the people for the job. In the end, we think that this new way of hiring people will make it easier for companies to find software engineers and will help people get jobs that they are really good at. It will also help reduce the costs of hiring people and make sure that everyone has a chance of getting a job.

Signaling Competence in Global Remote Technical Hiring: An Open-Source Framework