Generation Framework Research Articles

Implicit knowledge-augmented prompting for commonsense explanation generation

Abstract Commonsense explanation generation refers to reasoning and explaining why a commonsense statement contradicts commonsense knowledge, such as why the statement “My dad grew volleyballs in his garden” is nonsensical. While such reasoning is trivial for humans, it remains a challenge for AI systems. Despite their notable performance in tasks like text generation and reasoning, large language models (LLMs) often fall short of consistently generating coherent and accurate commonsense explanations. To bridge this gap, we propose a novel Two-stage Identification and Prompting (TIP) framework for enhancing LLMs’ ability to handle the task of commonsense explanation generation. Specifically, in the first stage, TIP identifies the nonsensical concept in the given statement, pinpointing the specific element that contradicts commonsense knowledge. In the second stage, TIP generates implicit knowledge based on the identified nonsensical concept and then leverages this implicit knowledge to guide the adopted LLMs in generating explanations. In order to demonstrate the effectiveness of the proposed TIP framework for commonsense explanation generation, we conducted extensive experiments based on the ComVE dataset and a newly constructed CSE dataset, where a variety of LLMs are evaluated. The experimental results show that TIP consistently outperforms all baseline methods across multiple metrics, demonstrating its effectiveness in improving LLMs’ commonsense reasoning and explanation generation capabilities.

Deep PackGen: A Deep Reinforcement Learning Framework for Adversarial Network Packet Generation

Recent advancements in artificial intelligence (AI) and machine learning (ML) algorithms, coupled with the availability of faster computing infrastructure, have enhanced the security posture of cybersecurity operations centers (defenders) through the development of ML-aided network intrusion detection systems (NIDS). Concurrently, the abilities of adversaries to evade security have also increased with the support of AI/ML models. Therefore, defenders need to proactively prepare for evasion attacks that exploit the detection mechanisms of NIDS. Recent studies have found that the perturbation of flow-based and packet-based features can deceive ML models, but these approaches have limitations. Perturbations made to the flow-based features are difficult to reverse-engineer, while samples generated with perturbations to the packet-based features are not playable. Our methodological framework, Deep PackGen, employs deep reinforcement learning to generate adversarial packets and aims to overcome the limitations of approaches in the literature. By taking raw malicious network packets as inputs and systematically making perturbations on them, Deep PackGen camouflages them as benign packets while still maintaining their functionality. In our experiments, using publicly available data, Deep PackGen achieved an average adversarial success rate of 66.4% against various ML models and across different attack types. Our investigation also revealed that more than 45% of the successful adversarial samples were out-of-distribution packets that evaded the decision boundaries of the classifiers. The knowledge gained from our study on the adversary’s ability to make specific evasive perturbations to different types of malicious packets can help defenders enhance the robustness of their NIDS against evolving adversarial attacks.

Leveraging natural language processing for automated assessment and feedback production in virtual education settings

With the rapid development of online teaching, the traditional teacher face-to-face evaluation and feedback model in online teaching faces problems such as high labor costs, low efficiency, and insufficient personalization. The application of natural language processing technology in the field of education has become a trend. Based on this choice, this article applied natural language processing to achieve automated evaluation and personalized feedback of online teaching processes. This article chose to use recurrent neural network (RNN) algorithm for text generation, and adopted Seq2Seq-based (Sequence-to-Sequence) recurrent neural network framework for automatic text feedback generation. Then, a group of student performance data was collected and divided into a control group and an experimental group, and comparisons were made between manual feedback and automatically generated feedback, respectively. Finally, the effectiveness and impact of automatic evaluation and feedback generation were evaluated by comparing the performance data and analysis results of two groups of students. The average accuracy of the automatic evaluation system designed in this article was 95.6%, which was higher than the 84.5% of manual evaluation. The accuracy of the automatic evaluation system designed in this article was higher, and the time spent on manual evaluation and automatic evaluation was compared. The average time spent on manual evaluation was 62.2 seconds, while automatic evaluation shortened the time to 36.98 seconds. For the feedback results, the graphical curves of the manually generated and automatically generated feedback results in terms of feedback on student grades were basically consistent, which can fully prove that the automatically generated feedback results have high accuracy. The remarkable efficacy of natural language processing (NLP) in online education is vividly illustrated by a recent study, where an NLP model notched an impressive average ROUGE score of 93.8 points. This achievement outstrips the performance of three other algorithmic models in the realm of automatic assessment and feedback generation. Such comparative experiments underscore the transformative potential of NLP technology, significantly boosting teaching efficiency and enriching the learning experience by adeptly tackling the perennial challenges of evaluation and feedback in online education settings.

A comparative study of AI‐generated and human‐crafted learning objectives in computing education

AbstractBackgroundIn computing education, educators are constantly faced with the challenge of developing new curricula, including learning objectives (LOs), while ensuring that existing courses remain relevant. Large language models (LLMs) were shown to successfully generate a wide spectrum of natural language artefacts in computing education.ObjectivesThe objective of this study is to evaluate if it is feasible for a state‐of‐the‐art LLM to support curricular design by proposing lists of high‐quality LOs.MethodsWe propose a simple LLM‐powered framework for the automatic generation of LOs. Two human evaluators compare the automatically generated LOs to the human‐crafted ones in terms of their alignment with course goals, meeting the SMART criteria, mutual overlap, and appropriateness of ordering.ResultsWe found that automatically generated LOs are comparable to LOs authored by instructors in many respects, including being measurable and relevant while exhibiting some limitations (e.g., sometimes not being specific or achievable). LOs were also comparable in their alignment with the high‐level course goals. Finally, auto‐generated LOs were often deemed to be better organised (order, non‐overlap) than the human‐authored ones.ConclusionsOur findings suggest that LLM could support educators in designing their courses by providing reasonable suggestions for LOs.

3DSMILES-GPT: 3D molecular pocket-based generation with token-only large language model.

The generation of three-dimensional (3D) molecules based on target structures represents a cutting-edge challenge in drug discovery. Many existing approaches often produce molecules with invalid configurations, unphysical conformations, suboptimal drug-like qualities, limited synthesizability, and require extensive generation times. To address these challenges, we present 3DSMILES-GPT, a fully language-model-driven framework for 3D molecular generation that utilizes tokens exclusively. We treat both two-dimensional (2D) and 3D molecular representations as linguistic expressions, combining them through full-dimensional representations and pre-training the model on a vast dataset encompassing tens of millions of drug-like molecules. This token-only approach enables the model to comprehensively understand the 2D and 3D characteristics of large-scale molecules. Subsequently, we fine-tune the model using pair-wise structural data of protein pockets and molecules, followed by reinforcement learning to further optimize the biophysical and chemical properties of the generated molecules. Experimental results demonstrate that 3DSMILES-GPT generates molecules that comprehensively outperform existing methods in terms of binding affinity, drug-likeness (QED), and synthetic accessibility score (SAS). Notably, it achieves a 33% enhancement in the quantitative estimation of QED, meanwhile the binding affinity estimated by Vina docking maintaining its state-of-the-art performance. The generation speed is remarkably fast, with the average time approximately 0.45 seconds per generation, representing a threefold increase over the fastest existing methods. This innovative 3DSMILES-GPT approach has the potential to positively impact the generation of 3D molecules in drug discovery.

Optimal operation strategies for freight transport with electric vehicles considering wireless charging lanes

Unlike traditional recharging at fixed stations, wireless charging lanes equipped on the road network enable electric vehicles (EVs) to recharge while in motion, reducing the time required for EV recharging. Indeed, the availability of wireless charging lanes allows the urban freight transport service operators to consider adjusting EVs routing plan to do EV recharging in transit so that the total operational costs can be further reduced. Envisioning that wireless charging lanes would be practically constructed on the road network, this study aims to investigate how the optimal EV routing plan can be determined in a pickup and delivery problem in the future electrified mobility system. The problem is formulated into a mixed-integer programming model, and we propose a branch-and-price algorithm to solve it. Additionally, a large neighborhood search heuristic is incorporated within the column generation framework to effectively tackle the sub-problems. Computational experiments on test instances highlight the effectiveness of our proposed algorithm and demonstrate the potential of wireless charging lanes to reduce total operational costs.

A novel digital unit cell library generation framework for topology optimization of multi-morphology lattice structures

A novel digital unit cell library generation framework for topology optimization of multi-morphology lattice structures

Novel Framework for Artificial Bubble Image Generation and Boundary Detection Using Superformula Regression and Computer Vision Techniques

Bubble multiphase systems are crucial in industries such as biotechnology, medicine, oil and gas, and water treatment. Optical data analysis provides critical insights into bubble characteristics, such as the shape and size, complementing physical sensor data. Existing detection techniques rely on classical computer vision algorithms and neural network models. While neural networks achieve a higher accuracy, they require extensive annotated datasets, and classical methods often struggle with complex systems due to their lower accuracy. This study proposes a novel framework to address these limitations. Using Superformula parameter regression, we introduce an advanced border detection method for accurately identifying gas inclusions and complex-shaped objects in multiphase environments. The framework also includes a new approach for generating realistic artificial bubble images based on physical flow conditions, leveraging the Superformula to create extensive, labeled datasets without manual annotation. Tested on real bubble flows in mass transfer equipment, the algorithms enable bubble classification by shape and size, enhance detection accuracy, and reduce development time for neural network solutions. This work provides a robust method for object detection and dataset generation in multiphase systems, paving the way for more precise modeling and analysis.

Leveraging domain knowledge for synthetic ultrasound image generation: a novel approach to rare disease AI detection.

This study explores the use of deep generative models to create synthetic ultrasound images for the detection of hemarthrosis in hemophilia patients. Addressing the challenge of sparse datasets in rare disease diagnostics, the study aims to enhance AI model robustness and accuracy through the integration of domain knowledge into the synthetic image generation process. The study employed two ultrasound datasets: a base dataset (Db) of knee recess distension images from non-hemophiliac patients and a target dataset (Dt) of hemarthrosis images from hemophiliac patients. The synthetic generation framework included a content generator (Gc) trained on Db and a context generator (Gs) to adapt these images to match Dt's context. This approach generated a synthetic target dataset (Ds), primed for AI training in rare disease research. The assessment of synthetic image generation involved expert evaluations, statistical analysis, and the use of domain-invariant perceptual distance and Fréchet inception distance for quality measurement. Expert evaluation revealed that images produced by our synthetic generation framework were comparable to real ones, with no significant difference in overall quality or anatomical accuracy. Additionally, the use of synthetic data in training convolutional neural networks demonstrated robustness in detecting hemarthrosis, especially with limited sample sizes. This study presents a novel approach for generating synthetic ultrasound images for rare disease detection, such as hemarthrosis in hemophiliac knees. By leveraging deep generative models and integrating domain knowledge, the proposed framework successfully addresses the limitations of sparse datasets and enhances AI model training and robustness. The synthetic images produced are of high quality and contribute significantly to AI-driven diagnostics in rare diseases, highlighting the potential of synthetic data in medical imaging.

ConoDL: a deep learning framework for rapid generation and prediction of conotoxins.

Conotoxins, being small disulfide-rich and bioactive peptides, manifest notable pharmacological potential and find extensive applications. However, the exploration of conotoxins' vast molecular space using traditional methods is severely limited, necessitating the urgent need of developing novel approaches. Recently, deep learning (DL)-based methods have advanced to the molecular generation of proteins and peptides. Nevertheless, the limited data and the intricate structure of conotoxins constrain the application of deep learning models in the generation of conotoxins. We propose ConoDL, a framework for the generation and prediction of conotoxins, comprising the end-to-end conotoxin generation model (ConoGen) and the conotoxin prediction model (ConoPred). ConoGen employs transfer learning and a large language model (LLM) to tackle the challenges in conotoxin generation. Meanwhile, ConoPred filters artificial conotoxins generated by ConoGen, narrowing down the scope for subsequent research. A comprehensive evaluation of the peptide properties at both sequence and structure levels indicates that the artificial conotoxins generated by ConoDL exhibit a certain degree of similarity to natural conotoxins. Furthermore, ConoDL has generated artificial conotoxins with novel cysteine scaffolds. Therefore, ConoDL may uncover new cysteine scaffolds and conotoxin molecules, facilitating further exploration of the molecular space of conotoxins and the discovery of pharmacologically active variants.

An Iterative Pseudo Label Generation framework for semi-supervised hyperspectral image classification using the Segment Anything Model

Hyperspectral image classification in remote sensing often encounters challenges due to limited annotated data. Semi-supervised learning methods present a promising solution. However, their performance is heavily influenced by the quality of pseudo labels. This limitation is particularly pronounced during the early stages of training, when the model lacks adequate prior knowledge. In this paper, we propose an Iterative Pseudo Label Generation (IPG) framework based on the Segment Anything Model (SAM) to harness structural prior information for semi-supervised hyperspectral image classification. We begin by using a small number of annotated labels as SAM point prompts to generate initial segmentation masks. Next, we introduce a spectral voting strategy that aggregates segmentation masks from multiple spectral bands into a unified mask. To ensure the reliability of pseudo labels, we design a spatial-information-consistency-driven loss function that optimizes IPG to adaptively select the most dependable pseudo labels from the unified mask. These selected pseudo labels serve as iterative point prompts for SAM. Following a suitable number of iterations, the resultant pseudo labels can be employed to enrich the training data for the classification model. Experiments conducted on the Indian Pines and Pavia University datasets demonstrate that even a simple 2D CNN based classification model trained with our generated pseudo labels significantly outperforms eight state-of-the-art hyperspectral image classification methods.

SLIDE: A Unified Mesh and Texture Generation Framework with Enhanced Geometric Control and Multi-view Consistency

SLIDE: A Unified Mesh and Texture Generation Framework with Enhanced Geometric Control and Multi-view Consistency

Empowering Nanostores for Competitiveness and Sustainable Communities in Emerging Countries: A Generative Artificial Intelligence Strategy Ideation Process

This exploratory study investigates Generative Artificial Intelligence’s (GenAI) use in strategy ideation for nanostores—i.e., small independent grocery retailers—to enhance their competitiveness while contributing to community sustainability. Nanostores, particularly in emerging countries, face intense competition and rapidly changing trends. These stores adopt various strategies by leveraging their proximity to consumers in neighbourhoods, resulting in different business configurations. While the existing literature highlights the broader nanostores’ functions, there is limited research on how they may develop comprehensive strategies to face their challenges. By employing a thing ethnography methodology, this work proposes GenAI thing interviewing—i.e., with ChatGPT 3.5 and Microsoft Copilot—through incremental prompting to explore potential strategy ideation and practices. Key findings suggest GenAI conversations can aid shopkeepers in strategy ideation through human-like written language, aligning with small business dynamics and structures. This proposition results in a GenAI ideation framework for strategy generation and definition. Moreover, this technology can enhance nanostore competitiveness and sustainability impact by enacting improved strategy practices in stakeholder engagements. Accordingly, this work’s main contribution underscores a GenAI-enabled conversational approach to facilitate nanostores’ strategy ideation and embedding in everyday business operations. Future work must address the limitations and further investigate GenAI’s influence on human understanding and technological creation, strategy ideation, adoption, and usability in nanostores.

A spatiotemporal style transfer algorithm for dynamic visual stimulus generation.

Understanding how visual information is encoded in biological and artificial systems often requires the generation of appropriate stimuli to test specific hypotheses, but available methods for video generation are scarce. Here we introduce the spatiotemporal style transfer (STST) algorithm, a dynamic visual stimulus generation framework that allows the manipulation and synthesis of video stimuli for vision research. We show how stimuli can be generated that match the low-level spatiotemporal features of their natural counterparts, but lack their high-level semantic features, providing a useful tool to study object recognition. We used these stimuli to probe PredNet, a predictive coding deep network, and found that its next-frame predictions were not disrupted by the omission of high-level information, with human observers also confirming the preservation of low-level features and lack of high-level information in the generated stimuli. We also introduce a procedure for the independent spatiotemporal factorization of dynamic stimuli. Testing such factorized stimuli on humans and deep vision models suggests a spatial bias in how humans and deep vision models encode dynamic visual information. These results showcase potential applications of the STST algorithm as a versatile tool for dynamic stimulus generation in vision science.

A Framework for Optimizing Deep Learning-Based Lane Detection and Steering for Autonomous Driving.

Improving the ability of autonomous vehicles to accurately identify and follow lanes in various contexts is crucial. This project aims to provide a novel framework for optimizing a self-driving vehicle that can detect lanes and steer accordingly. A virtual sandbox environment was developed in Unity3D that provides a semi-automated procedural road and driving generation framework for a variety of road scenarios. Four types of segments replicate actual driving situations by directing the car using strategically positioned waypoints. A training dataset thus generated was used to train a behavioral driving model that employs a convolutional neural network to detect the lane and ensure that the car steers autonomously to remain within lane boundaries. The model was evaluated on real-world driving footage from Comma.ai, exhibiting an autonomy of 77% in low challenge road conditions and of 66% on roads with sharper turns.

Durian: A Comprehensive Benchmark for Structure-Based 3D Molecular Generation.

Three-dimensional (3D) molecular generation models employ deep neural networks to simultaneously generate both topological representation and molecular conformations. Due to their advantages in utilizing the structural and interaction information on targets, as well as their reduced reliance on existing bioactivity data, these models have attracted widespread attention. However, limited training and testing data sets and the unexpected biases inherent in single evaluation metrics pose a significant challenge in comparing these models in practical settings. In this work, we proposed Durian, an evaluation framework for structure-based 3D molecular generation that incorporates protein-ligand data with experimental affinity and a comprehensive array of physicochemical and geometric metrics. The benchmark tasks encompass assessing the capability of models to reproduce the property distribution of training sets, generate molecules with rational distributions of drug-related properties, and exhibit potential high affinity toward given targets. Binding affinities were evaluated using three independent docking methods (QuickVina2, Surflex and Gnina) with both "Dock" and "Score" modes to reduce false positives arising from conformational searches or scoring functions. Specifically, we applied Durian to six 3D molecular generation methods: LiGAN, Pocket2Mol, DiffSBDD, SBDD, GraphBP, and SurfGen. While most methods demonstrated the ability to generate drug-like small molecules with reasonable physicochemical properties, they exhibited varying degrees of limitations in balancing novelty, structural rationality, and synthetic accessibility, thereby constraining their practical applications in drug discovery. Based on a total of 17 metrics, Durian highlights the importance of multiobjective optimization in 3D molecular generation methods. For instance, SurfGen and SBDD showed relatively comprehensive performance but could benefit from further improvements in molecular conformational rationality. Our evaluation framework is expected to provide meaningful guidance for the selection, optimization, and application of 3D generative models in practical drug design tasks.

Caterpillar-Inspired Multi-Gait Generation Method for Series-Parallel Hybrid Segmented Robot.

The body structures and motion stability of worm-like and snake-like robots have garnered significant research interest. Recently, innovative serial-parallel hybrid segmented robots have emerged as a fundamental platform for a wide range of motion modes. To address the hyper-redundancy characteristics of these hybrid structures, we propose a novel caterpillar-inspired Stable Segment Update (SSU) gait generation approach, establishing a unified framework for multi-segment robot gait generation. Drawing inspiration from the locomotion of natural caterpillars, the segments are modeled as rigid bodies with six degrees of freedom (DOF). The SSU gait generation method is specifically designed to parameterize caterpillar-like gaits. An inverse kinematics solution is derived by analyzing the forward kinematics and identifying the minimum lifting segment, framing the problem as a single-segment end-effector tracking task. Three distinct parameter sets are introduced within the SSU method to account for the stability of robot motion. These parameters, represented as discrete hump waves, are intended to improve motion efficiency during locomotion. Furthermore, the trajectories for each swinging segment are determined through kinematic analysis. Experimental results validate the effectiveness of the proposed SSU multi-gait generation method, demonstrating the successful traversal of gaps and rough terrain.

An Economic Impact Analysis of the Covid-19 Pandemic in the Nepalese Tourism Sector

Background: A considerable number of studies in the Nepalese context have revealed that Nepal’s tourism sector has been adversely affected by the Covid-19 pandemic; however, none of these studies have quantified these consequences in monetary terms. This assessment is expected to offer valuable insight for enhancing the resilience of tourism sector to future global disruptions and developing tailored policies to bolster the Nepalese tourism sector against external shocks. Objectives: This study addresses a critical gap in understanding the full economic ramifications of the Covid-19 pandemic in the Nepalese tourism sector by quantifying the primary, secondary and tertiary revenue losses. By providing a comprehensive assessment of the pandemic’s impact, this study aims to inform policymakers and industry stakeholders in developing tailored strategies for recovery and resilience. Methods: Secondary data are used in this study. It employs Stynes et al.’s (2000) revised money generation framework and the Keynesian macroeconomic multiplier approach to assess the actual and expected economic impacts of tourism activities in Nepal during the pandemic periods of 2020 and 2021. The actual value is determined using the actual tourism statistics, while the expected value is based on the targeted tourism statistics reported by the Ministry of Culture, Tourism, and Civil Aviation, Nepal. The difference between the two estimates is attributed to the economic impact of the Covid-19 pandemic in those years. Results: This study reveals an expected revenue decrease of 1.038 billion US dollars in 2020 and 1.309 billion US dollars in 2021. In addition, the tourism multiplier values are declining in Nepal over the observed years. Conclusion: This study provides two key conclusions. First, the Nepalese tourism sector is susceptible to travel restrictions. Second, tourism revenue is being drained from the local economy due to the increasing importation of merchandise and services to satisfy the Nepalese tourism sector’s demand.

Examining the Impact of Merger and Acquisition Perception on Customer Loyalty in the Banking Sector: Evidence from Structural Equation Modeling

Background: A considerable number of studies in the Nepalese context have revealed that Nepal’s tourism sector has been adversely affected by the Covid-19 pandemic; however, none of these studies have quantified these consequences in monetary terms. This assessment is expected to offer valuable insight for enhancing the resilience of tourism sector to future global disruptions and developing tailored policies to bolster the Nepalese tourism sector against external shocks. Objectives: This study addresses a critical gap in understanding the full economic ramifications of the Covid-19 pandemic in the Nepalese tourism sector by quantifying the primary, secondary and tertiary revenue losses. By providing a comprehensive assessment of the pandemic’s impact, this study aims to inform policymakers and industry stakeholders in developing tailored strategies for recovery and resilience. Methods: Secondary data are used in this study. It employs Stynes et al.’s (2000) revised money generation framework and the Keynesian macroeconomic multiplier approach to assess the actual and expected economic impacts of tourism activities in Nepal during the pandemic periods of 2020 and 2021. The actual value is determined using the actual tourism statistics, while the expected value is based on the targeted tourism statistics reported by the Ministry of Culture, Tourism, and Civil Aviation, Nepal. The difference between the two estimates is attributed to the economic impact of the Covid-19 pandemic in those years. Results: This study reveals an expected revenue decrease of 1.038 billion US dollars in 2020 and 1.309 billion US dollars in 2021. In addition, the tourism multiplier values are declining in Nepal over the observed years. Conclusion: This study provides two key conclusions. First, the Nepalese tourism sector is susceptible to travel restrictions. Second, tourism revenue is being drained from the local economy due to the increasing importation of merchandise and services to satisfy the Nepalese tourism sector’s demand.

Cutting the scotogenic loop: adding flavor to dark matter

We introduce a framework for hybrid neutrino mass generation, wherein scotogenic dark sector particles, including dark matter, are charged non-trivially under the A4 flavor symmetry. The spontaneous breaking of the A4 group to residual Z2 subgroup results in the “cutting” of the radiative loop. As a consequence the neutrinos acquire mass through the hybrid “scoto-seesaw” mass mechanism, combining aspects of both the tree-level seesaw and one-loop scotogenic mechanisms, with the residual Z2 subgroup ensuring the stability of the dark matter. The flavor symmetry also leads to several predictions including the normal ordering of neutrino masses and “generalized μ − τ reflection symmetry” in leptonic mixing. Additionally, it gives testable predictions for neutrinoless double beta decay and a lower limit on the lightest neutrino mass. Finally, A4 → Z2 breaking also leaves its imprint on the dark sector and ties it with the neutrino masses and mixing. The model allows only scalar dark matter, whose mass has a theoretical upper limit of ≲ 600 GeV, with viable parameter space satisfying all dark matter constraints, available only up to about 80 GeV. Conversely, fermionic dark matter is excluded due to constraints from the neutrino sector. Various aspects of this highly predictive framework can be tested in both current and upcoming neutrino and dark matter experiments.