PhD Candidate · Adelaide, Australia

Eagle W. H. Liang

AI research for Trust, Safety, Security & Privacy — teaching models to know what they don't know.

Australian Institute for Machine Learning (AIML) · SA Pathology
Adelaide University

w.liang@adelaide.edu.au Google Scholar GitHub
Portrait of Eagle W. H. Liang

// News

  1. Sep 2026
    Two calibration preprints released on arXiv — work on risk-conditioned calibration with small validation sets and reliability drift under score-preserving shift. SRTS-BCE ↗ · Reliability drift ↗
  2. Aug 2026
    Focal Calibration Loss preprint revised — version 3 is publicly available on arXiv. Read the preprint ↗
  3. May 2026
    Preprint on routing-conditional calibration — a diagnostic study of whether routing traces add reliable calibration evidence beyond prediction confidence. Read the preprint ↗
  4. Apr 2026
    Paper published at The Web Conference 2026 — “Lifting Manifolds to Mitigate Pseudo-Alignment in LLM4TS,” with Liangwei Nathan Zheng and co-authors. DOI ↗
  5. 2026
    TraffiX-MoE published in the ADMA 2025 proceedings — Springer’s 2026 proceedings record, pages 273–280. DOI ↗
  6. Nov 2025
    Two papers presented at CIKM 2025, Seoul — work on calibration for Kolmogorov–Arnold Networks and medical segmentation models.
  7. Sep 2025
    CIKM 2025 Travel Award — awarded by the CIKM 2025 Organizing Committee in recognition of accepted papers and contributions, supporting participation and presentation at CIKM 2025 in Seoul, Korea.
  8. Aug 2024
    FinSen dataset released — a financial news & sentiment dataset spanning 197 countries. Get it on GitHub ↗

// Publications Google Scholar ↗

2026 · Published

  • Lifting Manifolds to Mitigate Pseudo-Alignment in LLM4TS Liangwei Nathan Zheng, Wenhao Liang, Wei Emma Zhang, Miao Xu, Olaf Maennel, Weitong Chen
    The Web Conference 2026 · pp. 3764–3775 DOI arXiv
  • TraffiX-MoE: A Traffic-Aware Neural VRP Solver Wenhao Liang, Wei Emma Zhang, Lin Yue, Joy Rathjen, Peter O’Loughlin, Weitong Chen
    ADMA 2025 · Springer LNCS vol. 16200 · pp. 273–280 DOI

Publicly available preprints

  • Signal-Routed Temperature Scaling: Low-Capacity Risk-Conditioned Calibration for Small Validation Budgets Wenhao Liang, Liangwei Nathan Zheng, Lin Yue, Wei Emma Zhang, Mingyu Guo, Olaf Maennel, Weitong Chen
    arXiv preprint · Sep 2026 arXiv
  • How Much Can Reliability Drift Under a Fixed Confidence Distribution? Wenhao Liang, Lin Yue, Wei Emma Zhang, Mingyu Guo, Olaf Maennel, Weitong Chen
    arXiv preprint · Sep 2026 arXiv
  • Probing Routing-Conditional Calibration in Attention-Residual Transformers Wenhao Liang, Lin Yue, Wei Emma Zhang, Miao Xu, Mingyu Guo, Olaf Maennel, Weitong Chen
    arXiv preprint · May 2026 arXiv
  • Calibration Attention: Learning Reliability-Aware Representations for Vision Transformers Wenhao Liang, Wei Emma Zhang, Lin Yue, Miao Xu, Mingyu Guo, Olaf Maennel, Weitong Chen
    arXiv preprint · v2, Jan 2026 arXiv
  • Focal Calibration Loss: Controlling Posterior Distortion in Deep Neural Classifiers Wenhao Liang, Liangwei Zheng, Wei Zhang, Weitong Chen
    arXiv preprint · v3, Aug 2026 arXiv

2025 · Published

  • Calibrating on Kolmogorov-Arnold Network Wenhao Liang, Wei Emma Zhang, Lin Yue, Miao Xu, Olaf Maennel, Weitong Chen
    CIKM 2025 · pp. 1768–1777 DOI arXiv
  • Calibrating on Medical Segmentation Model through Signed Distance Wenhao Liang, Wei Emma Zhang, Lin Yue, Miao Xu, Olaf Maennel, Weitong Chen
    CIKM 2025 · pp. 1778–1787 DOI arXiv
  • Enhancing Financial Market Predictions: Causality-Driven Feature Selection Wenhao Liang, Zhengyang Li, Weitong Chen
    ADMA 2024 · proceedings published 2025 · pp. 149–163 DOI arXiv
  • Correlation Analysis of Adversarial Attack in Time Series Classification Zhengyang Li, Wenhao Liang, Chang Dong, Weitong Chen, Dong Huang
    ADMA 2024 · proceedings published 2025 · pp. 281–296 DOI arXiv

// Research

My research develops trustworthy and reliable AI, with a focus on medical diagnostics, model calibration, and evidence-based evaluation. I study how confidence, explanations, and model behaviour hold up across data, tasks, and deployment settings.

Trustworthy Medical AI

Reliable and interpretable machine learning for medical imaging and diagnostics, including segmentation and pathology-focused applications.

Probability Calibration

Training-time objectives and post-hoc methods that make predictive confidence more reliable, especially when validation data are limited.

Evidence and Causal Evaluation

Methods for examining whether AI-generated assessments are supported by relevant evidence and how conclusions change under controlled comparisons.

Robustness and Generalization

Evaluation of model reliability under distribution shift, across architectures, and in settings where errors carry practical consequences.

// Research Group — DT Lab

The Data Transpose (DT) Lab at Adelaide University focuses on AI, big data, and cloud-computing research aimed at transforming industries through cutting-edge technology. 📍 The University of Adelaide, North Terrace, Adelaide, SA

Lab Leader

A/Prof. Wei Emma Zhang

Dr. Wei Emma Zhang is an Associate Professor at the School of Computer Science and Information Technology and a researcher at the Australian Institute for Machine Learning (AIML), Adelaide University. She holds an ARC Early Career Industry Fellowship and is lead investigator of an ARC Discovery Project and an ARC Linkage Project. Author of 140+ publications across NLP, text mining, machine learning, and distributed systems, she is a recipient of the South Australia Young Tall Poppy Award (2024) and the Women of Colour in STEM Award (2025). She received her PhD in Computer Science from the University of Adelaide in 2017 and is an Honorary Lecturer at Macquarie University.

Lab Leader · DT-Lab Co-Director

Dr. Tony Weitong Chen

Dr. Tony Weitong Chen is a Senior Lecturer at Adelaide University, a core researcher at the Australian Institute for Machine Learning (AIML), and holder of the prestigious ARC Early Career Industry Fellowship. His research develops advanced machine-learning frameworks for medical data, multimodal information fusion, and knowledge discovery from imperfect data, with over $2.1 million in funding secured from the ARC, the Australia Economic Accelerator, and industry and government partners. He serves as Associate Editor of Neural Networks and Array, and earned his PhD from the University of Queensland in 2020, having previously been a Postdoctoral Research Fellow and Associate Lecturer at UQ.