AI 2025: Advances in Artificial Intelligence

This two-volume set LNAI 16370-16371 constitutes the refereed proceedings of the 38th Australasian Joint Conference on Artificial Intelligence, AI 2025, held in Canberra, ACT, Australia, during December 1 5, 2025.

The 73 full papers presented together with 1 short papers were carefully reviewed and selected from 152 submissions. They were organized in following topical section:

Part I : Language and Foundation Models and Knowledge Representation and Data Mining,

Part II : Learning Algorithms, Computer Vision, AI for Healthcare and Reinforcement Learning and Robotics.

. - Learning Algorithms.
. - Price Equilibrium Routing: A Lightweight Framework for Expert Selection in Mixture-of-Experts.
. - Minimum Message Length t-test.
. - Towards Scalable Backpropagation-Free Gradient Estimation.
. - Coordinate-free k-means clustering.
. - The Ensemble Kalman Update is an Empirical Matheron Update.
. - Using Individual Problem Instances for Exploratory Black Box Optimisation Benchmarking: A Case Study Using XOR Neural Networks.
. - On Knowledge-Informed Deep Learning for Modelling Complex Spatiotemporal Systems.
. - Trajectory segmentation and self-supervised classification for subtask discovery.
. - Computer Vision.
. - NoPo3DFusion: Scaling Two Images to Long Videos via 3D-Aware Iterative Diffusion.
. - Gaussian Alignment for Relative Camera Pose Estimation via Single-View Reconstruction.
. - Street Depth-Aware Gaussian for Modeling Dynamic Urban Scenes.
. - Paired Hierarchical VAEs for Image-to-Image Translation and Cross Reconstruction.
. - When Language Model Guides Vision: Grounding DINO for Cattle Muzzle Detection.
. - AFFT: Adapter-based Few-shot Fine-Tuning Framework for Remote Sensing Object Detection.
. - Depth-aware Audio Visual Segmentation with Geometry-Heuristic Cross Attention.
. - EyeDentify: A Deep Learning Approach to Non-Invasive Biometric Identification from Eye Blink Patterns.
. - Augmentation and Transformation for Nighttime Cloud Segmentation in All-Sky Camera Images.
. - Room Envelopes: A Synthetic Dataset for Indoor Layout Reconstruction from Images.
. - PLNet-12: A Vision-Language Benchmark for Zero-Shot Physical Literacy Analysis Across 12 Fundamental Movements.
. - TrashTracer: Enabling efficient real-time detection of underwater marine debris.
. - AI for Healthcare.
. - A-PriDiff: Anatomical Prior-Guided Conditional Diffusion for Ultrasound Spine Image Synthesis.
. - MammoMix: Leveraging Mixture of Experts for Robust Mammogram Breast Detection.
. - TripletResNet: A Deep Metric Learning Approach for mTBI Diagnosis from 3D CT.
. - Efficient Craniofacial Microsomia Detection via Edge-focused 3D Point Cloud Network
. - Privacy-Centric Seizure Detection Using Surface Normals, Pose and Segmentation Masks.
. - An Explainable Graph Learning Framework for Severe Maternal Morbidity Prediction.
. - Personalized Federated Graph Learning for Heterogeneous Incomplete EHRs.
. - Structure Matters: Brain Graph Augmentation via Learnable Edge Masking for Data-efficient Psychiatric Diagnosis.
. - Reinforcement Learning and Robotics.
. - Evolving MCTS Macro-Actions in Real-Time Domains.
. - A Hybrid Multi-Agent Reinforcement Learning Framework for Decentralised Search-And-Interact Tasks Under Partial Observability.
. - Synergistic MARL: Unifying Physics-Informed, Meta Learning, and Hybrid Learning to Outperform Communication-Based Coordination.
. - Guardrail Guided Policy Optimisation: Learning Disentangled Safety Constraints.
. - The Consensus Paradox: When Low Disagreement Leads to Catastrophic Failure in Multi-Teacher Reinforcement Learning.
. - Towards Search Node-Specific HTN Heuristics.
. - Planner-Independent Extraction of Goals and Constraints from Natural Language for Open-World Mobile Robot Missions.
. - PERCY: Personal Emotional Robotic Conversational System.
. - Understanding Human Situation Awareness in One-to-Many Human-Robot Interaction Scenarios.
. - Learning Preferences in Additive Separable Hedonic Project Games.
. - Policy Gradient Based Reinforcement Weighted Aggregation for Efficient Federated Learning on Heterogeneous Data.