This LNAI volume set constitutes the proceedings of 19th International Conference on Multi-disciplinary Trends in Artificial Intelligence, MIWAI 2026, held in Halifax, NS, Canada, during October 1012, 2026.

The 75 full papers included in these volumes were carefully reviewed and selected from 200 submissions. The conference has remained steadfastly dedicated to advancing high-quality research while embracing a wide, multi-disciplinary perspective on AI's diverse applications.

Inhalt

.- AURORA: Robust Interactive Left-Turn Planning at Unsignalized Intersections for Autonomous Vehicles with Uncertainty and Non-Holonomic Curvature Awareness.

.- Detector Selection as a Network Science Problem: AccuracyScalability Trade-offs in Distributed Edge Surveillance.

.- Multi-Scale Visual Feature Extraction for Medical Visual Question Answering.

.- Calibration-Aware Lightweight Model Selection for Trustworthy Skin Lesion Classification.

.- InjectGuard-RAG: Lightweight Defence Against Indirect Prompt Injection in Retrieval-Augmented Small-LLM Agents.

.- CART-Doc: A Compact Design Space for Confidence-Routed Document Key Information Extraction.

.- CMF-Lite: Lightweight Cross-Cue Fusion for Generalizable Deepfake Video Detection.

.- LiteMedRAG: Selective Retrieval-Augmented Compact Medical Visual Question Answering.

.- C-TAG: Context-aware Temporal Attention Graph for Trajectory Prediction.

.- Resource-Efficient Knowledge Graph Question Answering with Small Language Models.

.- Knowing When to Defer: A Risk-Calibrated Deferral Policy for Agentic AI in Financial Decision-Making.

.- PseuAgg: Enhancing Efficiency and Accuracy in Decentralized Federated Learning with Pseudo Aggregation.

.- Hybrid Semantic Positioning via Sim-to-Real Transfer: Leveraging Digital Twins for Constrained Edge Localization.

.- A Framework for Time-Series Anomaly Detection with Mamba Backbone.

.- SlimFusion: Lightweight AudioVisual Emotion Recognition for Edge Inference.

.- Sentiment-Aware Semantic Graph Construction for Harmful Tweet Classification using Graph Neural Networks.

.- Generative AIDriven Brain Tumor Segmentation with Counterfactual Explainability via Healthy Brain Reconstruction.

.- Hybrid Feature Selection with a Meta-Learner for Explainable IoT Intrusion Detection.

.- Provable Limits and Certified Deferral for Verbalized Uncertainty in Small Language Models.

.- When Identical Conditions Yield Different Authority: Measuring Path-Dependence in Trust-to-Authorization Policies.

.- LLM-Based Code Generation for Android Automotive OS Vehicle Hardware Abstraction Layer.

.- Multi-Level Knowledge Distillation for Lightweight Image Captioning.

.- SPFormer: A Physics-Guided Spectral Modeling with Adaptive Spectral Decomposition for Underwater Image Enhancement.

.- Solving the Faculty Timetabling Problem via the Quantum Approximate Optimization Algorithm.

.- Attentional Regression approach incorporating User-based and Review-based Explanations for Recommendation Systems.

.- NFTransUNet: A Neuro-Fuzzy Deep Learning Model for Breast Tumor Segmentation in Big Data Environment.

.- Placement Matters: An Accuracy-Cost Study of Attention Modules for Lightweight UAV Person Detection.

.- MORPH-YOLO: A High-Frequency Morphology-Preserving Neck for Fine-Grained Pest Detection.

.- Rank-Collapse and Detail Re-injection for Efficient Safety-Helmet Detection.

.- Symmetry Mappings of Causality Structures for Causal AI Formalised in a Galois Extension of Commutative Fields and Reified in a Noncommutative Field Extension.

.- A Monte Carlo Dropout Framework for Uncertainty-Aware Endoscopic Depth Estimation.

.- LIPE: Lightweight Landmark-Guided Image Patch Embedder for Real-Time Edge Gaze Estimation.

.- SACTTNet: A Hybrid Convolutional-Transformer Architecture with Spatially-Aware Convolutional Tokenization for Improved Bean Disease Classification.

.- Feature Refinement via C2f-Attention: An Enhanced YOLOv8 Approach for Tomato Ripeness Detection.

.- Enhancing YOLOv8 with Symmetric DINO for Low-Data Brain Tumor Detection.

.- ICU Readmission Prediction for Patients with Alcohol-related Disorders Based on Multimodal Deep Learning.

.- KAN-SegFormer: Kolmogorov-Arnold Decoding for Brain Tumor Segmentation.