Zhen Fang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zhen Fang has a total of 74 co-authors. Here are the top co-authors by number of shared publications:
- Jie Lu
University of Technology Sydney
23 shared publications - Guangquan Zhang
University of Technology Sydney
16 shared publications - Feng Liu
The University of Melbourne
11 shared publications - Bo Han
Hong Kong Baptist University
7 shared publications - Ling Chen
College of Chemistry Beijing Normal University Beijing P.R. China
4 shared publications - Tongliang Liu
The University of Sydney, Sydney, Australia
4 shared publications
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 34  (11) , 3745-3758
BTBR: A Bayesian-Theory-Driven Probabilistic–Fuzzy Framework for Implicit Bias Removal in Large Language Models
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 34  (11) , 3745-3758
BTBR: A Bayesian-Theory-Driven Probabilistic–Fuzzy Framework for Implicit Bias Removal in Large Language Models
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 48  (11) , 15112-15124
Feature-Alteration Robustness for Out-of-Distribution Detection
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 48  (11) , 15112-15124
Feature-Alteration Robustness for Out-of-Distribution Detection
- Open Access
- 9 September 2026
- Published by Elsevier in Computer Methods and Programs in Biomedicine
- Vol. 287
Predicting risk of ischemic stroke: A transformer model using genomic data
- 9 September 2026
- Published by Elsevier in Computer Methods and Programs in Biomedicine
- Vol. 287
Open AccessPredicting risk of ischemic stroke: A transformer model using genomic data
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 2802-2813
Training a Generalist Hallucination Detector across Multiple Domains via Adaptive Layer Aggregation
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 2802-2813
Open AccessTraining a Generalist Hallucination Detector across Multiple Domains via Adaptive Layer Aggregation
- Open Access
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 3647-3651
Partial Label Learning-Inspired Denoising Implicit Feedback for Recommendation
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 3647-3651
Open AccessPartial Label Learning-Inspired Denoising Implicit Feedback for Recommendation
- Open Access
- 15 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 2906-2907
Toward Trustworthy Vision-language Models in the Wild: Theory, Algorithm and Application
- 15 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 2906-2907
Open AccessToward Trustworthy Vision-language Models in the Wild: Theory, Algorithm and Application
- Open Access
- 20 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1171-1181
Debiased Negative Mining Improves Out-of-distribution Detection with Pre-trained Vision-Language Models
Aiming at identifying unexpected inputs from unknown classes, out-of-distribution (OOD) detection has emerged as a pivotal approach to enhancing the reliability of machine learning models. This paper focuses on the burgeoning paradigm of post-hoc OOD detection with pre-trained vision-language models (VLMs), where a popular pipeline is to detect OOD inputs by examining their affinities between ID labels and negative labels, i.e., those semantically different from ID labels. Due to the unavailability of target OOD labels, existing works predominantly rely on heuristic rules to mine negative labels from unlabeled wild corpus data. Despite the empirical success, we argue that the power of VLM-based OOD detection has yet to be fully unleashed since the notorious false negative problem is far from addressed in the literature. With this motivation, we are interested in addressing the challenge of mining true negative labels for OOD scoring. To this end, we develop a theoretical framework for correcting the sampling bias of negatives labels by indirectly approximating the distribution of negative labels. Perhaps surprisingly, we show that the debiased negative mining can be naturally converted into Monte-Carlo sampling based on ID labels and the unlabeled wild corpus data. Extensive experiments empirically manifest that our method establishes a new state-of-the-art in a variety of OOD detection setups.
- 20 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1171-1181
Open AccessDebiased Negative Mining Improves Out-of-distribution Detection with Pre-trained Vision-Language Models
Aiming at identifying unexpected inputs from unknown classes, out-of-distribution (OOD) detection has emerged as a pivotal approach to enhancing the reliability of machine learning models. This paper focuses on the burgeoning paradigm of post-hoc OOD detection with pre-trained vision-language models (VLMs), where a popular pipeline is to detect OOD inputs by examining their affinities between ID labels and negative labels, i.e., those semantically different from ID labels. Due to the unavailability of target OOD labels, existing works predominantly rely on heuristic rules to mine negative labels from unlabeled wild corpus data. Despite the empirical success, we argue that the power of VLM-based OOD detection has yet to be fully unleashed since the notorious false negative problem is far from addressed in the literature. With this motivation, we are interested in addressing the challenge of mining true negative labels for OOD scoring. To this end, we develop a theoretical framework for correcting the sampling bias of negatives labels by indirectly approximating the distribution of negative labels. Perhaps surprisingly, we show that the debiased negative mining can be naturally converted into Monte-Carlo sampling based on ID labels and the unlabeled wild corpus data. Extensive experiments empirically manifest that our method establishes a new state-of-the-art in a variety of OOD detection setups.
- 11 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Systems, Man, and Cybernetics: Systems
- Vol. 56  (6) , 3537-3549
Pseudo-Label Refinement for Multimodal Unsupervised Domain Adaptation
Unsupervised domain adaptation (UDA) aims to improve a model’s performance on an unlabeled target domain by leveraging labeled data from a source domain. Traditional UDA methods often overlook the rich textual information inherent in class labels, limiting their effectiveness. Recent advances in visual language models (VLMs), especially the contrastive language–image pretraining (CLIP) model, provide a promising multimodal foundation by integrating both image and text representations. However, most CLIP-based UDA approaches rely on pseudo-labels generated in a zero-shot manner by the pretrained CLIP model, which may become suboptimal as training progresses. To address this limitation, we propose a pseudo-label refinement (PuRe) framework that begins training with zero-shot CLIP pseudo-labels and then transitions to self-training, where we iteratively refine the model’s own predictions as pseudo-labels. In this iterative process, we introduce label consistency learning to stabilize predictions under strong data augmentations, increasing robustness, and an information maximization (IM) loss to encourage high-confidence predictions while preserving prediction diversity. Together, these components progressively enhance pseudo-label reliability, leading to improved adaptation in the target domain. Furthermore, we propose a novel prompt learning (PL) method that fine-tunes only a minimal set of parameters, specifically a single context token and class embeddings. Extensive experiments demonstrate that PuRe surpasses existing UDA methods on multiple benchmarks.
- 11 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Systems, Man, and Cybernetics: Systems
- Vol. 56  (6) , 3537-3549
Pseudo-Label Refinement for Multimodal Unsupervised Domain Adaptation
Unsupervised domain adaptation (UDA) aims to improve a model’s performance on an unlabeled target domain by leveraging labeled data from a source domain. Traditional UDA methods often overlook the rich textual information inherent in class labels, limiting their effectiveness. Recent advances in visual language models (VLMs), especially the contrastive language–image pretraining (CLIP) model, provide a promising multimodal foundation by integrating both image and text representations. However, most CLIP-based UDA approaches rely on pseudo-labels generated in a zero-shot manner by the pretrained CLIP model, which may become suboptimal as training progresses. To address this limitation, we propose a pseudo-label refinement (PuRe) framework that begins training with zero-shot CLIP pseudo-labels and then transitions to self-training, where we iteratively refine the model’s own predictions as pseudo-labels. In this iterative process, we introduce label consistency learning to stabilize predictions under strong data augmentations, increasing robustness, and an information maximization (IM) loss to encourage high-confidence predictions while preserving prediction diversity. Together, these components progressively enhance pseudo-label reliability, leading to improved adaptation in the target domain. Furthermore, we propose a novel prompt learning (PL) method that fine-tunes only a minimal set of parameters, specifically a single context token and class embeddings. Extensive experiments demonstrate that PuRe surpasses existing UDA methods on multiple benchmarks.
- Open Access
How Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability
Semantic associations such as the link between "bird" and "flew" are foundational for language modeling as they enable models to go beyond memorization and instead generalize and generate coherent text. Understanding how these associations are learned and represented in language models is essential for connecting deep learning with linguistic theory and developing a mechanistic foundation for large language models. In this work, we analyze how these associations emerge from natural language data in attention-based language models through the lens of training dynamics. By leveraging a leading-term approximation of the gradients, we develop closed-form expressions for the weights at early stages of training that explain how semantic associations first take shape. Through our analysis, we reveal that each set of weights of the transformer has closed-form expressions as simple compositions of three basis functions (bigram, token-interchangeability, and context mappings), reflecting the statistics of the text corpus and uncovering how each component of the transformer captures semantic associations based on these compositions. Experiments on real-world LLMs demonstrate that our theoretical weight characterizations closely match the learned weights, and qualitative analyses further show how our theorem shines light on interpreting the learned associations in transformers.
Open AccessHow Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability
Semantic associations such as the link between "bird" and "flew" are foundational for language modeling as they enable models to go beyond memorization and instead generalize and generate coherent text. Understanding how these associations are learned and represented in language models is essential for connecting deep learning with linguistic theory and developing a mechanistic foundation for large language models. In this work, we analyze how these associations emerge from natural language data in attention-based language models through the lens of training dynamics. By leveraging a leading-term approximation of the gradients, we develop closed-form expressions for the weights at early stages of training that explain how semantic associations first take shape. Through our analysis, we reveal that each set of weights of the transformer has closed-form expressions as simple compositions of three basis functions (bigram, token-interchangeability, and context mappings), reflecting the statistics of the text corpus and uncovering how each component of the transformer captures semantic associations based on these compositions. Experiments on real-world LLMs demonstrate that our theoretical weight characterizations closely match the learned weights, and qualitative analyses further show how our theorem shines light on interpreting the learned associations in transformers.
- Open Access
- 11 January 2026
- Published by Springer Nature in International Journal of Computer Vision
- Vol. 134  (2) , 54
Cross-domain Few-shot Classification via Invariant-content Feature Reconstruction
In cross-domain few-shot classification (CFC), mainstream studies aim to train a simple module (e.g. a linear transformation head) to select or transform features (a.k.a., the high-level semantic features) for previously unseen domains with a few labeled training data available on top of a powerful pre-trained model. These studies usually assume that high-level semantic features are shared across these domains, and just simple feature selection or transformations are enough to adapt features to previously unseen domains. However, in this paper, we find that the simply transformed features are too general to fully cover the key content features regarding each class. Thus, we propose an effective method, invariant-content feature reconstruction (IFR), to train a simple module that simultaneously considers both high-level and fine-grained invariant-content features for the previously unseen domains. Specifically, the fine-grained invariant-content features are considered as a set of informative and discriminative features learned from a few labeled training data of tasks sampled from unseen domains and are extracted by retrieving features that are invariant to style modifications from a set of content-preserving augmented data in pixel level with an attention module. Extensive experiments on the Meta-Dataset benchmark show that IFR achieves good generalization performance on unseen domains, which demonstrates the effectiveness of the fusion of the high-level features and the fine-grained invariant-content features. Specifically, IFR improves the average accuracy on unseen domains by 1.6% and 6.5% respectively under two different cross-domain few-shot classification settings.
- 11 January 2026
- Published by Springer Nature in International Journal of Computer Vision
- Vol. 134  (2) , 54
Open AccessCross-domain Few-shot Classification via Invariant-content Feature Reconstruction
In cross-domain few-shot classification (CFC), mainstream studies aim to train a simple module (e.g. a linear transformation head) to select or transform features (a.k.a., the high-level semantic features) for previously unseen domains with a few labeled training data available on top of a powerful pre-trained model. These studies usually assume that high-level semantic features are shared across these domains, and just simple feature selection or transformations are enough to adapt features to previously unseen domains. However, in this paper, we find that the simply transformed features are too general to fully cover the key content features regarding each class. Thus, we propose an effective method, invariant-content feature reconstruction (IFR), to train a simple module that simultaneously considers both high-level and fine-grained invariant-content features for the previously unseen domains. Specifically, the fine-grained invariant-content features are considered as a set of informative and discriminative features learned from a few labeled training data of tasks sampled from unseen domains and are extracted by retrieving features that are invariant to style modifications from a set of content-preserving augmented data in pixel level with an attention module. Extensive experiments on the Meta-Dataset benchmark show that IFR achieves good generalization performance on unseen domains, which demonstrates the effectiveness of the fusion of the high-level features and the fine-grained invariant-content features. Specifically, IFR improves the average accuracy on unseen domains by 1.6% and 6.5% respectively under two different cross-domain few-shot classification settings.