chatgpt

This study addresses the neglect of cognitive diversity among newcomers in open-source communities by mainstream large language models (LLMs). We propose the first AI-powered, personalized dialogue support paradigm grounded in problem-solving style (PSS). Our method comprises three core components: (1) a cognitive style modeling framework; (2) empirical analysis of newcomer interaction behaviors in open-source settings; and (3) a persona-based prompting engineering approach that dynamically adapts AI responses to users’ distinct reasoning patterns. By explicitly modeling PSS, our framework mitigates implicit stylistic biases inherent in generic LLMs—thereby enhancing fairness, accessibility, and pedagogical effectiveness of AI guidance. Key contributions include: (1) formal definition and empirical validation of PSS as a critical dimension for personalization; (2) a style-aware dialogue support framework integrating cognitive modeling and adaptive prompting; and (3) a scalable, interpretable AI empowerment pathway tailored to marginalized newcomer cohorts.

This study addresses the inefficiency of traditional programming exam design and its limited capacity to holistically assess students’ creativity, problem-solving skills, and domain knowledge. It presents the first systematic application of prompt engineering to the automatic generation of programming examination questions, proposing a method that leverages carefully crafted, diverse prompt templates to guide ChatGPT—without requiring fine-tuning of large language models. The approach autonomously produces high-quality questions and reference solutions spanning theoretical and practical aspects, multiple question types, and varying difficulty levels. Experimental results demonstrate that the generated items match or exceed the quality of human-authored questions while substantially improving item development efficiency. User studies further confirm the method’s effectiveness and practical value in educational settings.

This study evaluates the practical efficacy of GPT-4o in enhancing web accessibility compliance with WCAG 2.2, addressing the widespread lack of inclusive design in mainstream websites. Method: We introduce the first integrated approach combining multimodal inputs (including webpage screenshots), dynamic interaction testing, manual accessibility audits, and structured, context-aware prompt engineering—augmented by automated contrast analysis, semantic HTML validation, and compliance checking. Contribution/Results: Our novel “visual + interactive + semantic” tripartite feedback mechanism significantly improves detection and remediation of complex accessibility defects (e.g., focus management failures, ARIA logic errors). Experiments demonstrate that GPT-4o, guided by structured prompts and visual cues, efficiently resolves foundational issues and substantially increases overall WCAG compliance. However, it cannot yet replace human experts for end-to-end production workflows. This work establishes the first empirically grounded, reproducible framework for leveraging large language models to advance inclusive front-end development.

Visualization practitioners often lack formal training, resulting in significant gaps in design knowledge. Method: This paper presents the first systematic evaluation of large language models (LLMs) for visualization design consultation, employing a dual-path approach: quantitative analysis (multidimensional comparison of ChatGPT and expert responses from VisGuides) and qualitative analysis (double-blind user studies, content coding, and human-AI co-feedback analysis). Results: While ChatGPT rapidly generates diverse design alternatives, it substantially underperforms human experts in deep contextual understanding, visual intent inference, and support for nonlinear interactions. The study proposes a novel “context-enhanced + interaction-guided” paradigm tailored to design feedback, establishing both theoretical foundations and a practical framework for developing trustworthy LLM-assisted visualization design systems.

This study systematically evaluates ChatGPT’s capability boundaries and real-world applicability in NLP. Addressing the coexistence of its technical strengths, reliability deficiencies, and ethical risks, we propose the first multidimensional evaluation framework integrating technical performance, prompt engineering strategies, and ethical impact assessment. Through cross-model comparative experiments—spanning GPT-3 to GPT-4 Omni, LLaMA 3, and Gemini—we quantify state-of-the-art performance and critical failure modes across dialogue generation, content creation, machine translation, and clinical assistance tasks. We identify core limitations: bias amplification, factual hallucination, and insufficient safety alignment. Based on empirical findings, we provide deployable bias mitigation pathways and alignment optimization recommendations. Our work establishes a methodological foundation and practical guidance for enhancing large language model safety and domain-specific adaptation.

This study investigates whether prompt engineering training can improve secondary science teachers’ attitudes toward using ChatGPT as an instructional support tool. Method: Employing a quasi-experimental design, the study integrates pre- and post-intervention surveys with classroom implementation analysis to systematically examine— for the first time—the impact of targeted prompt engineering training on teachers’ technology acceptance. Contribution/Results: Training significantly enhanced teachers’ perceived usefulness, ease of use, and pedagogical appropriateness of ChatGPT (p < 0.01), alongside marked increases in usage intention and operational self-efficacy. The study’s key contribution lies in extending prompt engineering from a technical skill to a teacher professional development domain, thereby providing empirical evidence and a scalable, practice-informed training framework for the trustworthy and effective integration of large language models in K–12 education.

This study investigates how developers collaboratively integrate large language models (LLMs) into GitHub Pull Request (PR) workflows by sharing ChatGPT conversation links to facilitate code review and merging, specifically addressing: *What types of queries do developers pose to ChatGPT, and how do these influence their tangible contributions?* Method: We construct the first taxonomy of 14 query categories, grounded in empirical analysis of 155 real-world PRs containing embedded ChatGPT share links; coding and qualitative analysis are performed manually. Contribution/Results: As the first field study of human–LLM interaction within open-source development, our work systematically characterizes interaction types and collaboration patterns. Findings reveal that code review and task-oriented implementation queries dominate; code-generation requests involve significantly more conversational turns, whereas technical explanation and text polishing yield faster responses. The study elucidates the integration pathways and practical efficacy boundaries of LLMs in authentic software engineering workflows.

This study addresses the challenge of capturing authentic large language model (LLM) usage behaviors in real-world settings. We conducted the first empirical investigation among urban professionals in India, analyzing 238 unedited user prompts. Methodologically, we introduced the “retrospective anonymous social media survey,” integrating retrospective self-reports with qualitative–behavioral pattern analysis to overcome limitations of conventional self-reporting and controlled laboratory experiments. Results reveal that 85% of users engage daily, and 42.5% deeply integrate LLMs into professional workflows; ChatGPT is widely repurposed as a holistic life assistant spanning work, personal life, health, and creative domains. Users develop culturally adaptive prompting strategies, reflecting context-sensitive, locally grounded interaction practices. This research provides the first large-scale, ecologically valid evidence on AI adaptation mechanisms among Global South users, establishing a novel methodological paradigm for studying situated LLM use.

This study systematically evaluates ChatGPT (v3.5 and v4) on Japanese–English translation, benchmarking against leading commercial engines (e.g., Google Translate, DeepL) and distinguishing document-level versus sentence-level translation performance. It further investigates the impact of simple versus context-augmented prompting on output quality. Evaluation employs both automatic metrics (BLEU, COMET) and human annotation via the MQM framework. Results show: (1) Document-level context substantially improves coherence and coreference consistency; (2) GPT-3.5 prioritizes accuracy, whereas GPT-4 excels in fluency and stylistic appropriateness; (3) With optimized context-aware prompting, GPT-4 achieves parity with state-of-the-art commercial systems. This work provides the first empirical validation—specifically for Japanese–English—of LLMs’ document-level translation advantages and uncovers critical trade-offs between prompting strategies and model versions in translation quality.

This study investigates how prompt structure influences downstream collaborative integration behaviors in pull requests (PRs) during large language model (LLM)-assisted software development. Building on three dimensions—context, specificity, and verifiability—the authors develop an analytical framework for prompt structure and introduce a hybrid human–LLM annotation approach to address inconsistencies in LLM-based prompt evaluation. Empirical analysis reveals that specificity most effectively facilitates the generation of executable code, verifiability predominantly governs code adoption decisions, and context significantly affects integration depth. This work establishes, for the first time, a phased linkage between prompt structure and PR outcomes, uncovering the differential roles of prompt characteristics across distinct stages of the development process.