Yuanzhi Yao

Rui Wang, Yi Zheng, Dongxin Wang et al.

Existing topic modeling methods, from LDA to recent neural and LLM-based approaches, which focus mainly on statistical coherence, often produce redundant or off-target topics that miss the user's underlying intent. We introduce Human-centric Topic Modeling, \emph{Human-TM}), a novel task formulation that integrates a human-provided goal directly into the topic modeling process to produce interpretable, diverse and goal-oriented topics. To tackle this challenge, we propose the \textbf{G}oal-prompted \textbf{C}ontrastive \textbf{T}opic \textbf{M}odel with \textbf{O}ptimal \textbf{T}ransport (GCTM-OT), which first uses LLM-based prompting to extract goal candidates from documents, then incorporates these into semantic-aware contrastive learning via optimal transport for topic discovery. Experimental results on three public subreddit datasets show that GCTM-OT outperforms state-of-the-art baselines in topic coherence and diversity while significantly improving alignment with human-provided goals, paving the way for more human-centric topic discovery systems.

Yuting Ma, Yuanzhi Yao, Xiaohua Xu

Federated learning (FL) has attracted growing attention since it allows for privacy-preserving collaborative training on decentralized clients without explicitly uploading sensitive data to the central server. However, recent works have revealed that it still has the risk of exposing private data to adversaries. In this paper, we conduct reconstruction attacks and enhance inference attacks on various datasets to better understand that sharing trained classification model parameters to a central server is the main problem of privacy leakage in FL. To tackle this problem, a privacy-preserving image distribution sharing scheme with GAN (PPIDSG) is proposed, which consists of a block scrambling-based encryption algorithm, an image distribution sharing method, and local classification training. Specifically, our method can capture the distribution of a target image domain which is transformed by the block encryption algorithm, and upload generator parameters to avoid classifier sharing with negligible influence on model performance. Furthermore, we apply a feature extractor to motivate model utility and train it separately from the classifier. The extensive experimental results and security analyses demonstrate the superiority of our proposed scheme compared to other state-of-the-art defense methods. The code is available at https://github.com/ytingma/PPIDSG.

Muyang Zheng, Yuanzhi Yao, Changting Lin et al.

Despite efforts to align large language models (LLMs) with societal and moral values, these models remain susceptible to jailbreak attacks -- methods designed to elicit harmful responses. Jailbreaking black-box LLMs is considered challenging due to the discrete nature of token inputs, restricted access to the target LLM, and limited query budget. To address the issues above, we propose an effective method for jailbreaking black-box large language Models via Iterative Semantic Tuning, named MIST. MIST enables attackers to iteratively refine prompts that preserve the original semantic intent while inducing harmful content. Specifically, to balance semantic similarity with computational efficiency, MIST incorporates two key strategies: sequential synonym search, and its advanced version -- order-determining optimization. We conduct extensive experiments on two datasets using two open-source and four closed-source models. Results show that MIST achieves competitive attack success rate, relatively low query count, and fair transferability, outperforming or matching state-of-the-art jailbreak methods. Additionally, we conduct analysis on computational efficiency to validate the practical viability of MIST.