Author name cluster

Jun Wang 0018

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

5 papers

1 author row

ICML Conference 2025 Conference Paper

Adversarial Cooperative Rationalization: The Risk of Spurious Correlations in Even Clean Datasets

Wei Liu 0144
Zhongyu Niu
Lang Gao
Zhiying Deng
Jun Wang 0018
Haozhao Wang
Ruixuan Li 0001

This study investigates the self-rationalization framework constructed with a cooperative game, where a generator initially extracts the most informative segment from raw input, and a subsequent predictor utilizes the selected subset for its input. The generator and predictor are trained collaboratively to maximize prediction accuracy. In this paper, we first uncover a potential caveat: such a cooperative game could unintentionally introduce a sampling bias during rationale extraction. Specifically, the generator might inadvertently create an incorrect correlation between the selected rationale candidate and the label, even when they are semantically unrelated in the original dataset. Subsequently, we elucidate the origins of this bias using both detailed theoretical analysis and empirical evidence. Our findings suggest a direction for inspecting these correlations through attacks, based on which we further introduce an instruction to prevent the predictor from learning the correlations. Through experiments on six text classification datasets and two graph classification datasets using three network architectures (GRUs, BERT, and GCN), we show that our method significantly outperforms recent rationalization methods.

Details

ICML Conference 2025 Conference Paper

Learning Optimal Multimodal Information Bottleneck Representations

Qilong Wu
Yiyang Shao
Jun Wang 0018
Xiaobo Sun

Leveraging high-quality joint representations from multimodal data can greatly enhance model performance in various machine-learning based applications. Recent multimodal learning methods, based on the multimodal information bottleneck (MIB) principle, aim to generate optimal MIB with maximal task-relevant information and minimal superfluous information via regularization. However, these methods often set regularization weights in an ad hoc manner and overlook imbalanced task-relevant information across modalities, limiting their ability to achieve optimal MIB. To address this gap, we propose a novel multimodal learning framework, Optimal Multimodal Information Bottleneck (OMIB), whose optimization objective guarantees the achievability of optimal MIB by setting the regularization weight within a theoretically derived bound. OMIB further addresses imbalanced task-relevant information by dynamically adjusting regularization weights per modality, ensuring the inclusion of all task-relevant information. Moreover, we establish a solid information-theoretical foundation for OMIB’s optimization and implement it under the variational approximation framework for computational efficiency. Finally, we empirically validate the OMIB’s theoretical properties on synthetic data and demonstrate its superiority over the state-of-the-art benchmark methods in various downstream tasks.

Details

ECAI Conference 2024 Conference Paper

Adversarial Attack for Explanation Robustness of Rationalization Models

Yuankai Zhang 0002
Lingxiao Kong
Haozhao Wang
Ruixuan Li 0001
Jun Wang 0018
Yuhua Li 0003
Wei Liu 0144

Rationalization models, which select a subset of input text as rationale—crucial for humans to understand and trust predictions—have recently emerged as a prominent research area in eXplainable Artificial Intelligence (XAI). However, most of previous studies mainly focus on improving the quality of the rationale, ignoring its robustness to malicious attack. Specifically, whether the rationalization models can still generate high-quality rationale under the adversarial attack remains unknown. To explore this, this paper proposes UAT2E, which aims to undermine the explainability of rationalization models without altering their predictions, thereby eliciting distrust in these models from human users. UAT2E employs the gradient-based search on triggers and then inserts them into the original input to conduct both the non-target and target attack. Experimental results on five datasets reveal the vulnerability of rationalization models in terms of explanation, where they tend to select more meaningless tokens under attacks. Based on this, we make a series of recommendations for improving rationalization models in terms of explanation.

Details

ECAI Conference 2024 Conference Paper

PAGE: Parametric Generative Explainer for Graph Neural Network

Yang Qiu
Wei Liu 0144
Jun Wang 0018
Ruixuan Li 0001

This article introduces PAGE, a parameterized generative interpretive framework. PAGE is capable of providing faithful explanations for any graph neural network without necessitating prior knowledge or internal details. Specifically, we train the autoencoder to generate explanatory substructures by designing appropriate training strategy. Due to the dimensionality reduction of features in the latent space of the autoencoder, it becomes easier to extract causal features leading to the model’s output, which can be easily employed to generate explanations. To accomplish this, we introduce an additional discriminator to capture the causality between latent causal features and the model’s output. By designing appropriate optimization objectives, the well-trained discriminator can be employed to constrain the encoder in generating enhanced causal features. Finally, these features are mapped to substructures of the input graph through the decoder to serve as explanations. Compared to existing methods, PAGE operates at the sample scale rather than nodes or edges, eliminating the need for perturbation or encoding processes as seen in previous methods. Experimental results on both artificially synthesized and real-world datasets demonstrate that our approach not only exhibits the highest faithfulness and accuracy but also significantly outperforms baseline models in terms of efficiency.

Details

ECAI Conference 2020 Conference Paper

Fine-Grained Text Sentiment Transfer via Dependency Parsing

Lulu Xiao
Xiaoye Qu
Ruixuan Li 0001
Jun Wang 0018
Pan Zhou 0001
Yuhua Li 0003

Fine-grained sentiment transfer demands to edit an input sentence on a given sentiment intensity while preserving its content, which largely extends traditional binary sentiment transfer. Previous works on sentiment transfer usually attempt to learn latent content representation disentangled from sentiment. However, it is difficult to completely separate these two factors and it is also not necessary. In this paper, we propose a novel model that learns the latent representation without disentanglement and leverages sentiment intensity as input to decoder for fine-grained sentiment control. Moreover, aligned sentences with the same content but different sentiment intensities are usually unavailable. Due to the lack of parallel data, we construct pseudo-parallel sentences (i. e, sentences with similar content but different intensities) to relieve the burden of our model. In specific, motivated by the fact that the sentiment word (e. g. , “delicious”) has a close relationship with the non-sentiment context word (e. g. , “food”), we use dependency parsing to capture the dependency relationship. The pseudo-parallel sentences are produced by replacing the sentiment word with a new one according to the specific context word. Besides, the difference between pseudo-parallel sentences and generated sentences and other constraints are utilized to guide the model precisely revising sentiment. Experiments on the Yelp dataset show that our method substantially improves the degree of content preservation and sentiment accuracy and achieves state-of-the-art performance.

Details