papers

Publications (55)

cs.LG2022

Concurrent Adversarial Learning for Large-Batch Training

Yong Liu, Xiangning Chen, Minhao Cheng +2

Large-batch training has become a commonly used technique when training neural networks with a large number of GPU/TPU processors. As batch size increases, stochastic optimizers te…

cs.CV2024

MuLan: Multimodal-LLM Agent for Progressive and Interactive Multi-Object Diffusion

Sen Li, Ruochen Wang, Cho-Jui Hsieh +2

Existing text-to-image models still struggle to generate images of multiple objects, especially in handling their spatial positions, relative sizes, overlapping, and attribute bind…

cs.CL2024

MOSSBench: Is Your Multimodal Language Model Oversensitive to Safe Queries?

Xirui Li, Hengguang Zhou, Ruochen Wang +3

Humans are prone to cognitive distortions -- biased thinking patterns that lead to exaggerated responses to specific stimuli, albeit in very different contexts. This paper demonstr…

cs.LG2020

Self-Progressing Robust Training

Minhao Cheng, Pin-Yu Chen, Sijia Liu +3

Enhancing model robustness under new and even adversarial environments is a crucial milestone toward building trustworthy machine learning systems. Current robust training methods…

cs.CR2026

Optimizing Token Choice for Code Watermarking: An RL Approach

Zhimeng Guo, Huaisheng Zhu, Siyuan Xu +3

Protecting intellectual property on LLM-generated code necessitates effective watermarking systems that can operate within code's highly structured, syntactically constrained natur…

cs.AI2026

A Survey of Multimodal Mathematical Reasoning: From Perception, Alignment to Reasoning

Tianyu Yang, Sihong Wu, Yilun Zhao +6

Multimodal Mathematical Reasoning (MMR) has recently attracted increasing attention for its capability to solve mathematical problems involving both textual and visual modalities.…

cs.CV2025

CLIPErase: Efficient Unlearning of Visual-Textual Associations in CLIP

Tianyu Yang, Lisen Dai, Xiangqi Wang +3

Machine unlearning (MU) has gained significant attention as a means to remove specific data from trained models without requiring a full retraining process. While progress has been…

cs.LG2024

Uncovering, Explaining, and Mitigating the Superficial Safety of Backdoor Defense

Rui Min, Zeyu Qin, Nevin L. Zhang +2

Backdoor attacks pose a significant threat to Deep Neural Networks (DNNs) as they allow attackers to manipulate model predictions with backdoor triggers. To address these security…

cs.LG2023

Towards Stable Backdoor Purification through Feature Shift Tuning

Rui Min, Zeyu Qin, Li Shen +1

It has been widely observed that deep neural networks (DNN) are vulnerable to backdoor attacks where attackers could manipulate the model behavior maliciously by tampering with a s…

cs.AI2026

CARV: A Diagnostic Benchmark for Compositional Analogical Reasoning in Multimodal LLMs

Yongkang Du, Xiaohan Zou, Minhao Cheng +1

Analogical reasoning tests a fundamental aspect of human cognition: mapping the relation from one pair of objects to another. Existing evaluations of this ability in multimodal lar…

cs.CL2026

Route Before Retrieve: Activating Latent Routing Abilities of LLMs for RAG vs. Long-Context Selection

Yiwen Chen, Kuan Li, Fuzhen Zhuang +6

Recent advances in large language models (LLMs) have expanded the context window to beyond 128K tokens, enabling long-document understanding and multi-source reasoning. A key chall…

cs.LG2020

Voting based ensemble improves robustness of defensive models

Devvrit, Minhao Cheng, Cho-Jui Hsieh +1

Developing robust models against adversarial perturbations has been an active area of research and many algorithms have been proposed to train individual robust models. Taking thes…

cs.LG2020

CAT: Customized Adversarial Training for Improved Robustness

Minhao Cheng, Qi Lei, Pin-Yu Chen +2

Adversarial training has become one of the most effective methods for improving robustness of neural networks. However, it often suffers from poor generalization on both clean and…

cs.LG2025

Safety Reasoning with Guidelines

Haoyu Wang, Zeyu Qin, Li Shen +3

Training safe LLMs remains a critical challenge. The most widely used method, Refusal Training (RT), struggles to generalize against various Out-of-Distribution (OOD) jailbreaking…

cs.CV2024

The Crystal Ball Hypothesis in diffusion models: Anticipating object positions from initial noise

Yuanhao Ban, Ruochen Wang, Tianyi Zhou +3

Diffusion models have achieved remarkable success in text-to-image generation tasks; however, the role of initial noise has been rarely explored. In this study, we identify specifi…

cs.AI2025

R1-Zero's "Aha Moment" in Visual Reasoning on a 2B Non-SFT Model

Hengguang Zhou, Xirui Li, Ruochen Wang +3

Recently DeepSeek R1 demonstrated how reinforcement learning with simple rule-based incentives can enable autonomous development of complex reasoning in large language models, char…

cs.AI2025

Exploring Expert Failures Improves LLM Agent Tuning

Li-Cheng Lan, Andrew Bai, Minhao Cheng +2

Large Language Models (LLMs) have shown tremendous potential as agents, excelling at tasks that require multiple rounds of reasoning and interactions. Rejection Sampling Fine-Tunin…

cs.LG2018

Query-Efficient Hard-label Black-box Attack:An Optimization-based Approach

Minhao Cheng, Thong Le, Pin-Yu Chen +3

We study the problem of attacking a machine learning model in the hard-label black-box setting, where no model information is revealed except that the attacker can make queries to…

cs.LG2022

Efficient Non-Parametric Optimizer Search for Diverse Tasks

Ruochen Wang, Yuanhao Xiong, Minhao Cheng +1

Efficient and automated design of optimizers plays a crucial role in full-stack AutoML systems. However, prior methods in optimizer search are often limited by their scalability, g…

cs.LG2022

FedDM: Iterative Distribution Matching for Communication-Efficient Federated Learning

Yuanhao Xiong, Ruochen Wang, Minhao Cheng +2

Federated learning~(FL) has recently attracted increasing attention from academia and industry, with the ultimate goal of achieving collaborative training under privacy and communi…

cs.LG2026

Sparse MeZO: Less Parameters for Better Performance in Zeroth-Order LLM Fine-Tuning

Yong Liu, Zirui Zhu, Chaoyu Gong +3

While fine-tuning large language models (LLMs) for specific tasks often yields impressive results, it comes at the cost of memory inefficiency due to back-propagation in gradient-b…

cs.CR2024

DrAttack: Prompt Decomposition and Reconstruction Makes Powerful LLM Jailbreakers

Xirui Li, Ruochen Wang, Minhao Cheng +2

The safety alignment of Large Language Models (LLMs) is vulnerable to both manual and automated jailbreak attacks, which adversarially trigger LLMs to output harmful content. Howev…

cs.LG2021

RANK-NOSH: Efficient Predictor-Based Architecture Search via Non-Uniform Successive Halving

Ruochen Wang, Xiangning Chen, Minhao Cheng +2

Predictor-based algorithms have achieved remarkable performance in the Neural Architecture Search (NAS) tasks. However, these methods suffer from high computation costs, as trainin…

cs.CL2026

Scalable Token-Level Hallucination Detection in Large Language Models

Rui Min, Tianyu Pang, Chao Du +2

Large language models (LLMs) have demonstrated remarkable capabilities, but they still frequently produce hallucinations. These hallucinations are difficult to detect in reasoning-…

stat.ML2019

Enhancing Certifiable Robustness via a Deep Model Ensemble

Huan Zhang, Minhao Cheng, Cho-Jui Hsieh

We propose an algorithm to enhance certified robustness of a deep model ensemble by optimally weighting each base model. Unlike previous works on using ensembles to empirically imp…

cs.CV2024

Understanding the Impact of Negative Prompts: When and How Do They Take Effect?

Yuanhao Ban, Ruochen Wang, Tianyi Zhou +3

The concept of negative prompts, emerging from conditional generation models like Stable Diffusion, allows users to specify what to exclude from the generated images.%, demonstrati…

cs.CL2025

LaRA: Benchmarking Retrieval-Augmented Generation and Long-Context LLMs -- No Silver Bullet for LC or RAG Routing

Kuan Li, Liwen Zhang, Yong Jiang +4

Effectively incorporating external knowledge into Large Language Models (LLMs) is crucial for enhancing their capabilities and addressing real-world needs. Retrieval-Augmented Gene…

cs.IR2019

Natural Adversarial Sentence Generation with Gradient-based Perturbation

Yu-Lun Hsieh, Minhao Cheng, Da-Cheng Juan +3

This work proposes a novel algorithm to generate natural language adversarial input for text classification models, in order to investigate the robustness of these models. It invol…

cs.CL2023

Backdoor Learning on Sequence to Sequence Models

Lichang Chen, Minhao Cheng, Heng Huang

Backdoor learning has become an emerging research area towards building a trustworthy machine learning system. While a lot of works have studied the hidden danger of backdoor attac…

cs.CV2023

Attacking by Aligning: Clean-Label Backdoor Attacks on Object Detection

Yize Cheng, Wenbin Hu, Minhao Cheng

Deep neural networks (DNNs) have shown unprecedented success in object detection tasks. However, it was also discovered that DNNs are vulnerable to multiple kinds of attacks, inclu…

cs.LG2018

Towards Robust Neural Networks via Random Self-ensemble

Xuanqing Liu, Minhao Cheng, Huan Zhang +1

Recent studies have revealed the vulnerability of deep neural networks: A small adversarial perturbation that is imperceptible to human can easily make a well-trained deep neural n…

stat.ML2018

Stochastic Zeroth-order Optimization via Variance Reduction method

Liu Liu, Minhao Cheng, Cho-Jui Hsieh +1

Derivative-free optimization has become an important technique used in machine learning for optimizing black-box models. To conduct updates without explicitly computing gradient, m…

cs.CL2025

Scaling Laws of Synthetic Data for Language Models

Zeyu Qin, Qingxiu Dong, Xingxing Zhang +10

Large language models (LLMs) achieve strong performance across diverse tasks, largely driven by high-quality web data used in pre-training. However, recent studies indicate this da…

cs.LG2025

Input Snapshots Fusion for Scalable Discrete-Time Dynamic Graph Neural Networks

QingGuo Qi, Hongyang Chen, Minhao Cheng +1

In recent years, there has been a surge in research on dynamic graph representation learning, primarily focusing on modeling the evolution of temporal-spatial patterns in real-worl…

cs.CL2025

Improving Your Model Ranking on Chatbot Arena by Vote Rigging

Rui Min, Tianyu Pang, Chao Du +3

Chatbot Arena is a popular platform for evaluating LLMs by pairwise battles, where users vote for their preferred response from two randomly sampled anonymous models. While Chatbot…

cs.LG2025

POME: Post Optimization Model Edit via Muon-style Projection

Yong Liu, Di Fu, Yang Luo +4

We introduce Post-Optimization Model Edit (POME), a new algorithm that enhances the performance of fine-tuned large language models using only their pretrained and fine-tuned check…

cs.CR2025

PISanitizer: Preventing Prompt Injection to Long-Context LLMs via Prompt Sanitization

Runpeng Geng, Yanting Wang, Chenlong Yin +3

Long context LLMs are vulnerable to prompt injection, where an attacker can inject an instruction in a long context to induce an LLM to generate an attacker-desired output. Existin…

cs.CL2022

MSDT: Masked Language Model Scoring Defense in Text Domain

Jaechul Roh, Minhao Cheng, Yajun Fang

Pre-trained language models allowed us to process downstream tasks with the help of fine-tuning, which aids the model to achieve fairly high accuracy in various Natural Language Pr…

cs.LG2020

Sign-OPT: A Query-Efficient Hard-label Adversarial Attack

Minhao Cheng, Simranjit Singh, Patrick Chen +3

We study the most practical problem setup for evaluating adversarial robustness of a machine learning system with limited access: the hard-label black-box attack setting for genera…

cs.CL2026

ReSum: Unlocking Long-Horizon Search Intelligence via Context Summarization

Xixi Wu, Kuan Li, Yida Zhao +13

Large Language Model (LLM)-based web agents excel at knowledge-intensive tasks but face a fundamental conflict between the need for extensive exploration and the constraints of lim…

cs.CR2025

TASO: Jailbreak LLMs via Alternative Template and Suffix Optimization

Yanting Wang, Runpeng Geng, Jinghui Chen +2

Many recent studies showed that LLMs are vulnerable to jailbreak attacks, where an attacker can perturb the input of an LLM to induce it to generate an output for a harmful questio…

cs.LG2021

Rethinking Architecture Selection in Differentiable NAS

Ruochen Wang, Minhao Cheng, Xiangning Chen +2

Differentiable Neural Architecture Search is one of the most popular Neural Architecture Search (NAS) methods for its search efficiency and simplicity, accomplished by jointly opti…

cs.LG2020

Seq2Sick: Evaluating the Robustness of Sequence-to-Sequence Models with Adversarial Examples

Minhao Cheng, Jinfeng Yi, Pin-Yu Chen +2

Crafting adversarial examples has become an important technique to evaluate the robustness of deep neural networks (DNNs). However, most existing works focus on attacking the image…

cs.CR2024

A Watermark-Conditioned Diffusion Model for IP Protection

Rui Min, Sen Li, Hongyang Chen +1

The ethical need to protect AI-generated content has been a significant concern in recent years. While existing watermarking strategies have demonstrated success in detecting synth…

cs.CL2024

Defense Against Syntactic Textual Backdoor Attacks with Token Substitution

Xinglin Li, Xianwen He, Yao Li +1

Textual backdoor attacks present a substantial security risk to Large Language Models (LLM). It embeds carefully chosen triggers into a victim model at the training stage, and make…

cs.AI2026

Reason Before You Retrieve: Agentic Planning for Multi-modal RAG

Tianyu Yang, Shir Simon, Zhenzhen Li +2

Multimodal retrieval-augmented generation (mRAG) aims to answer image-text queries with external knowledge, but most existing systems still retrieve directly from raw multimodal in…

cs.LG2020

Attack Graph Convolutional Networks by Adding Fake Nodes

Xiaoyun Wang, Minhao Cheng, Joe Eaton +2

In this paper, we study the robustness of graph convolutional networks (GCNs). Previous work have shown that GCNs are vulnerable to adversarial perturbation on adjacency or feature…

cs.AI2024

One Prompt is not Enough: Automated Construction of a Mixture-of-Expert Prompts

Ruochen Wang, Sohyun An, Minhao Cheng +3

Large Language Models (LLMs) exhibit strong generalization capabilities to novel tasks when prompted with language instructions and in-context demos. Since this ability sensitively…

cs.LG2024

Invisible Backdoor Attacks on Diffusion Models

Sen Li, Junchi Ma, Minhao Cheng

In recent years, diffusion models have achieved remarkable success in the realm of high-quality image generation, garnering increased attention. This surge in interest is parallele…

cs.LG2021

DrNAS: Dirichlet Neural Architecture Search

Xiangning Chen, Ruochen Wang, Minhao Cheng +2

This paper proposes a novel differentiable architecture search method by formulating it into a distribution learning problem. We treat the continuously relaxed architecture mixing…

cs.CV2023

Dataset Distillation via Adversarial Prediction Matching

Mingyang Chen, Bo Huang, Junda Lu +4

Dataset distillation is the technique of synthesizing smaller condensed datasets from large original datasets while retaining necessary information to persist the effect. In this p…

cs.CR2021

A Review of Adversarial Attack and Defense for Classification Methods

Yao Li, Minhao Cheng, Cho-Jui Hsieh +1

Despite the efficiency and scalability of machine learning systems, recent studies have demonstrated that many classification methods, especially deep neural networks (DNNs), are v…

cs.LG2026

Empowering Reliable Visual-Centric Instruction Following in MLLMs

Weilei He, Feng Ju, Zhiyuan Fan +3

Evaluating the instruction-following (IF) capabilities of Multimodal Large Language Models (MLLMs) is essential for rigorously assessing how faithfully model outputs adhere to user…

cs.LG2023

Revisiting Personalized Federated Learning: Robustness Against Backdoor Attacks

Zeyu Qin, Liuyi Yao, Daoyuan Chen +3

In this work, besides improving prediction accuracy, we study whether personalization could bring robustness benefits to backdoor attacks. We conduct the first study of backdoor at…

cs.CL2023

PTP: Boosting Stability and Performance of Prompt Tuning with Perturbation-Based Regularizer

Lichang Chen, Heng Huang, Minhao Cheng

Recent studies show that prompt tuning can better leverage the power of large language models than fine-tuning on downstream natural language understanding tasks. However, the exis…