Author name cluster

Chenhang Cui

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.

13 papers

2 author rows

ICLR Conference 2025 Conference Paper

Fine-Grained Verifiers: Preference Modeling as Next-token Prediction in Vision-Language Alignment

Chenhang Cui
An Zhang 0003
Yiyang Zhou
Zhaorun Chen
Gelei Deng
Huaxiu Yao
Tat-Seng Chua

The recent advancements in large language models (LLMs) and pre-trained vision models have accelerated the development of vision-language large models (VLLMs), enhancing the interaction between visual and linguistic modalities. Despite their notable success across various domains, VLLMs face challenges in modality alignment, which can lead to issues like hallucinations and unsafe content generation. Current alignment techniques often rely on coarse feedback and external datasets, limiting scalability and performance. In this paper, we propose FiSAO (Fine-Grained Self-Alignment Optimization), a novel self-alignment method that utilizes the model’s own visual encoder as a fine-grained verifier to improve vision-language alignment without the need for additional data. By leveraging token-level feedback from the vision encoder, FiSAO significantly improves vision-language alignment, even surpassing traditional preference tuning methods that require additional data. Through both theoretical analysis and experimental validation, we demonstrate that FiSAO effectively addresses the misalignment problem in VLLMs, marking the first instance of token-level rewards being applied to such models. Our code is avaliable at \url{https://anonymous.4open.science/r/FISAO-57F0/}.

Details

NeurIPS Conference 2025 Conference Paper

MJ-Bench: Is Your Multimodal Reward Model Really a Good Judge for Text-to-Image Generation?

Zhaorun Chen
Zichen Wen
Yichao Du
Yiyang Zhou
Chenhang Cui
Siwei Han
Jen Weng
Chaoqi Wang

While text-to-image models like GPT-4o-Image and FLUX are rapidly proliferating, they often encounter challenges such as hallucination, bias, and the production of unsafe, low-quality output. To effectively address these issues, it is crucial to align these models with desired behaviors based on feedback from a multimodal judge. Despite their significance, current multimodal judges frequently undergo inadequate evaluation of their capabilities and limitations, potentially leading to misalignment and unsafe fine-tuning outcomes. To address this issue, we introduce MJ-Bench, a novel benchmark which incorporates a comprehensive preference dataset to evaluate multimodal judges in providing feedback for image generation models across six key perspectives: alignment, safety, image quality, bias, composition, and visualization. Specifically, we evaluate a large variety of multimodal judges including smaller-sized CLIP-based scoring models, open-source VLMs, and close-source VLMs on each decomposed subcategory of our preference dataset. Experiments reveal that close-source VLMs generally provide better feedback, with GPT-4o outperforming other judges in average. Compared with open-source VLMs, smaller-sized scoring models can provide better feedback regarding text-image alignment and image quality, while VLMs provide more accurate feedback regarding safety and generation bias due to their stronger reasoning capabilities. Further studies in feedback scale reveal that VLM judges can generally provide more accurate and stable feedback in natural language than numerical scales. Notably, human evaluations on end-to-end and fine-tuned models using separate feedback from these multimodal judges provide similar conclusions, further confirming the effectiveness of MJ-Bench.

PDF Details

ICLR Conference 2025 Conference Paper

MMIE: Massive Multimodal Interleaved Comprehension Benchmark for Large Vision-Language Models

Peng Xia 0005
Siwei Han
Shi Qiu 0016
Yiyang Zhou
Zhaoyang Wang 0004
Wenhao Zheng
Zhaorun Chen
Chenhang Cui

Interleaved multimodal comprehension and generation, enabling models to produce and interpret both images and text in arbitrary sequences, have become a pivotal area in multimodal learning. Despite significant advancements, the evaluation of this capability remains insufficient. Existing benchmarks suffer from limitations in data scale, scope, and evaluation depth, while current evaluation metrics are often costly or biased, lacking in reliability for practical applications. To address these challenges, we introduce MMIE, a large-scale knowledge-intensive benchmark for evaluating interleaved multimodal comprehension and generation in Large Vision-Language Models (LVLMs). MMIE comprises 20K meticulously curated multimodal queries, spanning 3 categories, 12 fields, and 102 subfields, including mathematics, coding, physics, literature, health, and arts. It supports both interleaved inputs and outputs, offering a mix of multiple-choice and open-ended question formats to evaluate diverse competencies. Moreover, we propose a reliable automated evaluation metric, leveraging a scoring model fine-tuned with human-annotated data and systematic evaluation criteria, aimed at reducing bias and improving evaluation accuracy. Extensive experiments demonstrate the effectiveness of our benchmark and metrics in providing a comprehensive evaluation of interleaved LVLMs. Specifically, we evaluate eight LVLMs, revealing that even the best models show significant room for improvement, with most achieving only moderate results. We believe MMIE will drive further advancements in the development of interleaved LVLMs.

Details

TMLR Journal 2025 Journal Article

Reliable and Responsible Foundation Models

Xinyu Yang
Junlin Han
Rishi Bommasani
Jinqi Luo
Wenjie Qu
Wangchunshu Zhou
Adel Bibi
Xiyao Wang

Foundation models, including Large Language Models (LLMs), Multimodal Large Language Models (MLLMs), Image Generative Models (i.e, Text-to-Image Models and Image-Editing Models), and Video Generative Models, have become essential tools with broad applications across various domains such as law, medicine, education, finance, and beyond. As these models see increasing real-world deployment, ensuring their reliability and responsibility has become critical for academia, industry, and government. This survey addresses the reliable and responsible development of foundation models. We explore critical issues, including bias and fairness, security and privacy, uncertainty, explainability, and distribution shift. Our research also covers model limitations, such as hallucinations, as well as methods like alignment and Artificial Intelligence-Generated Content (AIGC) detection. For each area, we review the current state of the field and outline concrete future research directions. Additionally, we discuss the intersections between these areas, highlighting their connections and shared challenges. We hope our survey fosters the development of foundation models that are not only powerful but also ethical, trustworthy, reliable, and socially responsible.

PDF Details

NeurIPS Conference 2025 Conference Paper

RSafe: Incentivizing proactive reasoning to build robust and adaptive LLM safeguards

jingnan zheng
Xiangtian Ji
Yijun Lu
Chenhang Cui
Weixiang Zhao
Gelei Deng
Zhenkai Liang
An Zhang

Large Language Models (LLMs) continue to exhibit vulnerabilities despite deliberate safety alignment efforts, posing significant risks to users and society. To safeguard against the risk of policy-violating content, system-level moderation via external guard models—designed to monitor LLM inputs and outputs and block potentially harmful content—has emerged as a prevalent mitigation strategy. Existing approaches of training guard models rely heavily on extensive human curated datasets and struggle with out-of-distribution threats, such as emerging harmful categories or jailbreak attacks. To address these limitations, we propose RSafe, an adaptive reasoning-based safeguard that conducts guided safety reasoning to provide robust protection within the scope of specified safety policies. RSafe operates in two stages: (1) guided reasoning, where it analyzes safety risks of input content through policy-guided step-by-step reasoning, and (2) reinforced alignment, where rule-based RL optimizes its reasoning paths to align with accurate safety prediction. This two-stage training paradigm enables RSafe to internalize safety principles to generalize safety protection capability over unseen or adversarial safety violation scenarios. During inference, RSafe accepts user-specified safety policies to provide enhanced safeguards tailored to specific safety requirements. Experiments demonstrate that RSafe matches state-of-the-art guard models using limited amount of public data in both prompt- and response-level harmfulness detection, while achieving superior out-of-distribution generalization on both emerging harmful category and jailbreak attacks. Furthermore, RSafe provides human-readable explanations for its safety judgments for better interpretability. RSafe offers a robust, adaptive, and interpretable solution for LLM safety moderation, advancing the development of reliable safeguards in dynamic real-world environments. Our code is available at https: //anonymous. 4open. science/r/RSafe-996D.

PDF Details

NeurIPS Conference 2025 Conference Paper

Safe + Safe = Unsafe? Exploring How Safe Images Can Be Exploited to Jailbreak Large Vision-Language Models

Chenhang Cui
Gelei Deng
An Zhang
jingnan zheng
Yicong Li
Lianli Gao
Tianwei Zhang
Tat-Seng Chua

Recent advances in Large Vision-Language Models (LVLMs) have showcased strong reasoning abilities across multiple modalities, achieving significant breakthroughs in various real-world applications. Despite this great success, the safety guardrail of LVLMs may not cover the unforeseen domains introduced by the visual modality. Existing studies primarily focus on eliciting LVLMs to generate harmful responses via carefully crafted image-based jailbreaks designed to bypass alignment defenses. In this study, we reveal that a safe image can be exploited to achieve the same jailbreak consequence when combined with additional safe images and prompts. This stems from two fundamental properties of LVLMs: universal reasoning capabilities and safety snowball effect. Building on these insights, we propose Safety Snowball Agent (SSA), a novel agent-based framework leveraging agents' autonomous and tool-using abilities to jailbreak LVLMs. SSA operates through two principal stages: (1) initial response generation, where tools generate or retrieve jailbreak images based on potential harmful intents, and (2) harmful snowballing, where refined subsequent prompts induce progressively harmful outputs. Our experiments demonstrate that SSA can use nearly any image to induce LVLMs to produce unsafe content, achieving high success jailbreaking rates against the latest LVLMs. Unlike prior works that exploit alignment flaws, SSA leverages the inherent properties of LVLMs, presenting a profound challenge for enforcing safety in generative multimodal systems.

PDF Details

AAAI Conference 2024 Conference Paper

Adaptive Feature Imputation with Latent Graph for Deep Incomplete Multi-View Clustering

Jingyu Pu
Chenhang Cui
Xinyue Chen
Yazhou Ren
Xiaorong Pu
Zhifeng Hao
Philip S. Yu
Lifang He

In recent years, incomplete multi-view clustering (IMVC), which studies the challenging multi-view clustering problem on missing views, has received growing research interests. Previous IMVC methods suffer from the following issues: (1) the inaccurate imputation for missing data, which leads to suboptimal clustering performance, and (2) most existing IMVC models merely consider the explicit presence of graph structure in data, ignoring the fact that latent graphs of different views also provide valuable information for the clustering task. To overcome such challenges, we present a novel method, termed Adaptive feature imputation with latent graph for incomplete multi-view clustering (AGDIMC). Specifically, it captures the embbedded features of each view by incorporating the view-specific deep encoders. Then, we construct partial latent graphs on complete data, which can consolidate the intrinsic relationships within each view while preserving the topological information. With the aim of estimating the missing sample based on the available information, we utilize an adaptive imputation layer to impute the embedded feature of missing data by using cross-view soft cluster assignments and global cluster centroids. As the imputation progresses, the portion of complete data increases, contributing to enhancing the discriminative information contained in global pseudo-labels. Meanwhile, to alleviate the negative impact caused by inferior impute samples and the discrepancy of cluster structures, we further design an adaptive imputation strategy based on the global pseudo-label and the local cluster assignment. Experimental results on multiple real-world datasets demonstrate the effectiveness of our method over existing approaches.

PDF Details DOI

ICLR Conference 2024 Conference Paper

Analyzing and Mitigating Object Hallucination in Large Vision-Language Models

Yiyang Zhou
Chenhang Cui
Jaehong Yoon
Linjun Zhang
Zhun Deng
Chelsea Finn
Mohit Bansal
Huaxiu Yao

Large vision-language models (LVLMs) have shown remarkable abilities in understanding visual information with human languages. However, LVLMs still suffer from object hallucination, which is the problem of generating descriptions that include objects that do not actually exist in the images. This can negatively impact many vision-language tasks, such as visual summarization and reasoning. To address this issue, we propose a simple yet powerful algorithm, LVLM Hallucination Revisor (LURE), to post-hoc rectify object hallucination in LVLMs by reconstructing less hallucinatory descriptions. LURE is grounded in a rigorous statistical analysis of the key factors underlying object hallucination, including co-occurrence (the frequent appearance of certain objects alongside others in images), uncertainty (objects with higher uncertainty during LVLM decoding), and object position (hallucination often appears in the later part of the generated text). LURE can also be seamlessly integrated with any LVLMs. We evaluate LURE on six open-source LVLMs and found it outperforms the previous best approach in both general object hallucination evaluation metrics, GPT, and human evaluations.

Details

NeurIPS Conference 2024 Conference Paper

Calibrated Self-Rewarding Vision Language Models

Yiyang Zhou
Zhiyuan Fan
Dongjie Cheng
Sihan Yang
Zhaorun Chen
Chenhang Cui
Xiyao Wang
Yun Li

Large Vision-Language Models (LVLMs) have made substantial progress by integrating pre-trained large language models (LLMs) and vision models through instruction tuning. Despite these advancements, LVLMs often exhibit the hallucination phenomenon, where generated text responses appear linguistically plausible but contradict the input image, indicating a misalignment between image and text pairs. This misalignment arises because the model tends to prioritize textual information over visual input, even when both the language model and visual representations are of high quality. Existing methods leverage additional models or human annotations to curate preference data and enhance modality alignment through preference optimization. These approaches are resource-intensive and may not effectively reflect the target LVLM's preferences, making the curated preferences easily distinguishable. Our work addresses these challenges by proposing the Calibrated Self-Rewarding (CSR) approach, which enables the model to self-improve by iteratively generating candidate responses, evaluating the reward for each response, and curating preference data for fine-tuning. In the reward modeling, we employ a step-wise strategy and incorporate visual constraints into the self-rewarding process to place greater emphasis on visual input. Empirical results demonstrate that CSR significantly enhances performance and reduces hallucinations across twelve benchmarks and tasks, achieving substantial improvements over existing methods by 7. 62\%. Our empirical results are further supported by rigorous theoretical analysis, under mild assumptions, verifying the effectiveness of introducing visual constraints into the self-rewarding paradigm. Additionally, CSR shows compatibility with different vision-language models and the ability to incrementally improve performance through iterative fine-tuning.

PDF Details DOI

IJCAI Conference 2024 Conference Paper

Dynamic Weighted Graph Fusion for Deep Multi-View Clustering

Yazhou Ren
Jingyu Pu
Chenhang Cui
Yan Zheng
Xinyue Chen
Xiaorong Pu
Lifang He

By exploring complex graph information hidden in data from multiple views, multi-view clustering based on graph neural network significantly enhances the clustering performance and has drawn increasing attention in recent years. Although considerable progress has been made, most existing GNN based MVC models merely consider the explicit presence of graph structure in raw data and ignore that latent graphs of different views also provide specific information for the clustering task. We propose dynamic weighted graph fusion for deep multi-view clustering (DFMVC) to address this issue. Specifically, DFMVC learns embedded features via deep autoencoders and then constructs latent graphs for each individual view. Then, it concatenates the embedded features of all views to form a global feature to leverage complementary information, as well as generates a fusion graph via combining all latent graphs to accurately capture the topological information among samples. Based on the informative fusion graph and global features, the graph convolution module is adopted to derive a representation with global comprehensive information, which is further used to generate pseudo-label information. In a self-supervised manner, such information guides each view to dynamically learn discriminative features and latent graphs. Extensive experimental results demonstrate the efficacy of DFMVC.

PDF Details DOI

IJCAI Conference 2024 Conference Paper

Integrating Vision-Language Semantic Graphs in Multi-View Clustering

JunLong Ke
Zichen Wen
Yechenhao Yang
Chenhang Cui
Yazhou Ren
Xiaorong Pu
Lifang He

In recent years, a variety of graph learning-based multi-view clustering (MVC) methods have emerged. However, these methods continue to face challenges in extracting latent features from real-world data, particularly in scenarios involving high-resolution color images and high-dimensional features. This task is notably difficult in cases where images are visually similar yet semantically diverse. To address this issue, we present a novel large-scale pre-trained model for multi-view clustering, named Integrate Vision-Language Semantic Graphs in Multi-View Clustering (IVSGMV), which harnesses the capabilities of visual-language pre-training models to enhance clustering performance and confronts issues in the unsupervised tuning of pre-trained models for multi-view data. We introduce an effective unsupervised approach for creating semantic graphs from image multi-view datasets using pre-trained encoders. Our method addresses the inherent spatial noise and imbalance in these encoders by employing graph filters and a joint process that integrates both image node and edge features. Additionally, we demonstrate the application of our approach to multi-view image clustering on extensive datasets, notably the high-resolution MVImgNet, achieving an impressive 82% accuracy. Furthermore, our method extends the zero-shot capabilities of large-scale pre-trained models, resulting in good performance in clustering tasks on untrained multi-view datasets.

PDF Details DOI

NeurIPS Conference 2023 Conference Paper

A Novel Approach for Effective Multi-View Clustering with Information-Theoretic Perspective

Chenhang Cui
Yazhou Ren
Jingyu Pu
Jiawei Li
Xiaorong Pu
Tianyi Wu
Yutao Shi
Lifang He

Multi-view clustering (MVC) is a popular technique for improving clustering performance using various data sources. However, existing methods primarily focus on acquiring consistent information while often neglecting the issue of redundancy across multiple views. This study presents a new approach called Sufficient Multi-View Clustering (SUMVC) that examines the multi-view clustering framework from an information-theoretic standpoint. Our proposed method consists of two parts. Firstly, we develop a simple and reliable multi-view clustering method SCMVC (simple consistent multi-view clustering) that employs variational analysis to generate consistent information. Secondly, we propose a sufficient representation lower bound to enhance consistent information and minimise unnecessary information among views. The proposed SUMVC method offers a promising solution to the problem of multi-view clustering and provides a new perspective for analyzing multi-view data. To verify the effectiveness of our model, we conducted a theoretical analysis based on the Bayes Error Rate, and experiments on multiple multi-view datasets demonstrate the superior performance of SUMVC.

PDF Details

IJCAI Conference 2023 Conference Paper

Deep Multi-view Subspace Clustering with Anchor Graph

Chenhang Cui
Yazhou Ren
Jingyu Pu
Xiaorong Pu
Lifang He

Deep multi-view subspace clustering (DMVSC) has recently attracted increasing attention due to its promising performance. However, existing DMVSC methods still have two issues: (1) they mainly focus on using autoencoders to nonlinearly embed the data, while the embedding may be suboptimal for clustering because the clustering objective is rarely considered in autoencoders, and (2) existing methods typically have a quadratic or even cubic complexity, which makes it challenging to deal with large-scale data. To address these issues, in this paper we propose a novel deep multi-view subspace clustering method with anchor graph (DMCAG). To be specific, DMCAG firstly learns the embedded features for each view independently, which are used to obtain the subspace representations. To significantly reduce the complexity, we construct an anchor graph with small size for each view. Then, spectral clustering is performed on an integrated anchor graph to obtain pseudo-labels. To overcome the negative impact caused by suboptimal embedded features, we use pseudo-labels to refine the embedding process to make it more suitable for the clustering task. Pseudo-labels and embedded features are updated alternately. Furthermore, we design a strategy to keep the consistency of the labels based on contrastive learning to enhance the clustering performance. Empirical studies on real-world datasets show that our method achieves superior clustering performance over other state-of-the-art methods.

PDF Details DOI