Hao Dong

hao.dong@ibk.baug.ethz.ch

I received my Ph.D. from ETH Zurich, co-advised by Prof. Eleni Chatzi and Prof. Olga Fink. I received my Master's degree in Control, Robotics, and Autonomous Systems from Aalto University in 2022. During my Master's, I also spent time at TUM as an exchange student, at CVG through the ETH Robotics Student Fellowship, and at StachnissLab, where I completed my Master's thesis. I received my Bachelor's degree from Xi’an Jiaotong University in 2020.

My long-term research vision is to advance a new generation of multimodal intelligent systems, characterized by robustness, reliability, and adaptability.

I'm always open for discussions or collaborations, feel free to get in touch via email.

Selected Publications

Towards Reliable Multimodal Learning for Real-world Systems

Hao Dong

Doctoral Thesis at ETH Zurich, 2026

thesis

Adaptive Confidence Regularization for Multimodal Failure Detection

Moru Liu, Hao Dong, Olga Fink, Mario Trapp

CVPR, 2026

Hao Dong

Selected Publications

Towards Reliable Multimodal Learning for Real-world Systems

Adaptive Confidence Regularization for Multimodal Failure Detection

Towards Multimodal Domain Generalization with Few Labels

Advances in Multimodal Adaptation and Generalization: From Traditional Approaches to Foundation Models

Adapting Vision-Language Models Without Labels: A Comprehensive Survey

To Trust Or Not To Trust Your Vision-Language Model's Prediction

Extremely Simple Multimodal Outlier Synthesis for Out-of-Distribution Detection and Segmentation

Towards Robust Multimodal Open-set Test-time Adaptation via Adaptive Entropy-aware Optimization

MultiOOD: Scaling Out-of-Distribution Detection for Multiple Modalities

Towards Multimodal Open-Set Domain Generalization and Adaptation through Self-supervision

SimMMDG: A Simple and Effective Framework for Multi-modal Domain Generalization

NNG-Mix: Improving Semi-supervised Anomaly Detection with Pseudo-anomaly Generation

Unseen Visual Anomaly Generation

DPU: Dynamic Prototype Updating for Multimodal Out-of-Distribution Detection

Interactive Symbolic Regression through Offline Reinforcement Learning: A Co-Design Framework

Recall and Refine: A Simple but Effective Source-free Open-set Domain Adaptation Framework

SuperFusion: Multilevel LiDAR-Camera Fusion for Long-Range HD Map Generation

Learning-Based Dimensionality Reduction for Computing Compact and Effective Local Feature Descriptors

Online Pole Segmentation on Range Images for Long-term LiDAR Localization in Urban Environments

Online Range Image-based Pole Extractor for Long-term LiDAR Localization in Urban Environments

Invited Talks

Towards Reliable Multimodal Learning for Real-world Systems

Multi-modal Fusion and Generalization

Robust Robot Perception: From Single-modality to Multimodality

Academic Services