AI MDPI

🚀We’re proud to share that AI has reached a 2025 Impact Factor of 6.5 (🎉+30%） This milestone reflects the vibrant contributions of our authors, reviewers, editors and readers—thank you! #ImpactFactor #AI 📢 Submit your AI research & grow with us! 👉 mdpi.com/journal/ai

📊 Latest #AI Research – Highlights 🛒 In a 12-seller/12-buyer GPT-4o-mini market, private history drove far more false quality claims than public history (0.850 vs 0.483). Information architecture shapes deceptive selling. Read: doi.org/10.3390/ai7090370 #LLM #MultiAgent

📊 Latest #AI Research – Highlights 🧠 ChatGPT, Grok, Gemini & DeepSeek on cognitive tests: all aced language, ChatGPT & Gemini beat university students in math, all lagged on visual–spatial. #LLM Read: doi.org/10.3390/ai7090371

📖 Top-cited #AI2024 (Open Access): "AI and Face-Driven Orthodontics: A Scoping Review of Digital Advances in Diagnosis & Treatment Planning" 🔗 doi.org/10.3390/ai5010009 From 2D/3D cephalometrics to facial analysis & treatment tracking — AI reshapes orthodontics. 62 citations 📊

📊 Latest #AI Research – Highlights 🔬 A CNN + Transformer network with cross-channel attention fuses HR 2D and light-field data for 3D measurement of micro-structured surfaces—beating current depth-estimation methods. #DeepLearning Read: doi.org/10.3390/ai7090374

📊 Latest #AI Research – Highlights 🦾 Augmenting IMU orientation data via musculoskeletal simulation boosts exercise-evaluation models—better accuracy, generalization to unseen subjects, and few-shot patient fine-tuning. #AIinHealth Read: doi.org/10.3390/ai7090381

📖 Top-cited #AI2024 (Open Access): "Secure Internet Financial Transactions: A Framework Integrating MFA & ML" 🔗 doi.org/10.3390/ai5010010 2-layer security: fingerprint + OTP → fraud triggers face recognition. LR hits 97.94% acc. 64 citations 📊 #FinTech #CyberSecurity

📊 Latest #AI Research – Highlights ⚡ Approximate multipliers usually hurt DNN accuracy. This study flips it: layer-wise INT8 approximation cut energy up to 44.75% while improving Top-1 accuracy by up to 2.07 pts. #DNN #Efficiency Read: doi.org/10.3390/ai7090385

📊 Latest #AI Research – Highlights 🔧 Can LLMs generate SysML-v2 artefacts for MBSE? Testing 4 LLMs × 5 prompt strategies: no single strategy won across tasks. Useful as a human-in-the-loop assistant—expert validation still essential. #MBSE #LLM Read: doi.org/10.3390/ai7080274

📊 Latest #AI Research – Highlights Why does AV AI lag behind LLMs? New in @AI_MDPI: Digital AI thrived on shared ecosystems; AV research stays fragmented, limiting reproducibility & scale. #AutonomousVehicles 🔗 doi.org/10.3390/ai7080275

📊 Latest #AI Research – Highlights "Let's Code with GenAI" — a self-paced PD module helping K-12 teachers tackle text-based coding with a GenAI assistant. 34 teachers: self-efficacy, value beliefs & coding/CT scores all improved. #AIinEducation #K12 🔗 doi.org/10.3390/ai7080277

📖 Top-cited #AI2024 (Open Access): "Trustworthy AI: Securing Sensitive Data in Large Language Models" 🔗 doi.org/10.3390/ai5040134 RBAC + ABAC + NER + differential privacy for dynamic output control. 77 citations 📊 Must-read for #LLMsecurity & #PrivacyAI 👇 RT to share!

📊 Latest #AI Research – Highlights 📚 Can semantic clustering automate few-shot exemplar selection for LLM math feedback? Feasibility study in AI 2026: 96% of instructor ratings were usable (3–5). Promising, not yet proven superior. #AIinEducation Read doi.org/10.3390/ai7080280

📊 Latest #AI Research – Highlights 🚨 Moderate optimization pressure—not high—produced the riskiest manipulation signals in AI marketing agents. Non-monotonic effects in an OpenClaw simulation. Pressure-aware auditing urged. #AIEthics Read: doi.org/10.3390/ai7080281

📖 Top-cited #AI2024 (Open Access): "xLSTMTime: Long-Term Time Series Forecasting 🔗 doi.org/10.3390/ai5030071 Exponential gating + revised memory beats Transformers & LTSF-Linear on real datasets. 84 citations 📊 Must-read for #TimeSeries & #LSTM builders 👇 RT to share!

📊 Latest #AI Research – Highlights Maskless object removal: a dual-pipeline framework (SAM–SDXL & LaMa) deletes a target object using only an ID, box, or point — no hand-painted mask required. More accurate and up to 32× faster. 🎨#ImageEditing 📄 doi.org/10.3390/ai7080284

📖 Top-cited #AI2024 (Open Access): "Strategic Approaches to Quantum AI Integration for Industry 4.0" 🔗 doi.org/10.3390/ai5010015 Change mgmt, R&D strategy & real cases from IBM, Google, VW, JPMorgan. 92 citations 📊 Must-read for #QuantumAI leaders 👇 RT to share!

📊 Latest #AI Research – Highlights Can deep learning predict TMS motor responses without endless sampling? 🧠 Flattened cortical E-field maps + CNNs hit ROC-AUC up to 0.87 — toward faster TMS motor mapping. 📄 doi.org/10.3390/ai7080285 #TMS #DeepLearning #Neuroimaging

📖 Top-cited #AI2024 (Open Access): "Error Correction and Adaptation in Conversational AI" 🔗 doi.org/10.3390/ai5020041 Feedback loops, RL, meta-learning + case studies across 6 domains. 98 citations 📊 Must-read for #Chatbot & #NLP builders 👇 RT to share!

📊 Latest #AI Research – Highlights Can local open-weight LLMs be trusted as safety judges? 🧑‍⚖️ SafetyJudge-LLM audits 7 models across 10,612 outputs and finds high detection ≠ low false alarms. Reliability needs multi-dimensional auditing. #LLMSafety 📄 doi.org/10.3390/ai7080286