11 papers
Suppressing Pressure, Amplifying Evidence: Self-Guided Attention Steering to Mitigate Sycophancy and Stubbornness
Yinghao He, Mengyu Xu, Haixiang Sun +8
Reliable language models should resist unsupported user pressure while effectively using objective contextual information. However, models may exhibit sycophancy by yielding to uns…
Positive-Unlabeled Learning for Agent Safety False Alarm Auditing
Xichen Yan, Chongyang Gao, Kezhen Chen +3
Safety monitors help safeguard language-model agents interacting with external tools and environments, but conservative monitoring can generate many false alarms, consuming extensi…
On Unlearning for Time-series Forecasting
Zeyu Shi, Yanhui Luo, Ziming Hong +4
Time-series forecasting is widely used in sensitive domains. Models in these settings are often trained on longitudinal user- or entity-level records, which may later require remov…
NüshuVoice: Reviving the Voice of Endangered Nüshu with Pitch-Aware Text-to-Speech
Hongkun Yang, Xinhui Yi, Xiyan Zhao +13
Nüshu is an endangered phonetic script historically used by women in Jiangyong County, southern Hunan, China. While existing computational studies of Nüshu mainly focus on textual…
Defense against Poisoning Attacks under Shuffle-DP
Siyi Wang, Qiyao Luo, Yihua Hu +6
Differential Privacy (DP) has become the gold standard for protecting individual privacy in data analytics, and the shuffle-DP model has attracted significant attention from both a…
DP-MGTD: Privacy-Preserving Machine-Generated Text Detection via Adaptive Differentially Private Entity Sanitization
Lionel Z. Wang, Yusheng Zhao, Jiabin Luo +6
The deployment of Machine-Generated Text (MGT) detection systems necessitates processing sensitive user data, creating a fundamental conflict between authorship verification and pr…
AppellateGen: A Benchmark for Appellate Legal Judgment Generation
Hongkun Yang, Lionel Z. Wang, Wei Fan +10
Legal judgment generation is a critical task in legal intelligence. However, existing research in legal judgment generation has predominantly focused on first-instance trials, rely…
The Man Behind the Sound: Demystifying Audio Private Attribute Profiling via Multimodal Large Language Model Agents
Lixu Wang, Kaixiang Yao, Xinfeng Li +4
Our research uncovers a novel privacy risk associated with multimodal large language models (MLLMs): the ability to infer sensitive personal attributes from audio data -- a techniq…
Token Buncher: Shielding LLMs from Harmful Reinforcement Learning Fine-Tuning
Weitao Feng, Lixu Wang, Peizhuo Lv +5
As large language models (LLMs) continue to grow in capability, so do the risks of harmful misuse through fine-tuning. While most prior studies assume that attackers rely on superv…
The Eye of Sherlock Holmes: Uncovering User Private Attribute Profiling via Vision-Language Model Agentic Framework
Feiran Liu, Yuzhe Zhang, Xinyi Huang +9
Our research reveals a new privacy risk associated with the vision-language model (VLM) agentic framework: the ability to infer sensitive attributes (e.g., age and health informati…
Shallow Flow Matching for Coarse-to-Fine Text-to-Speech Synthesis
Dong Yang, Yiyi Cai, Yuki Saito +2
We propose Shallow Flow Matching (SFM), a novel mechanism that enhances flow matching (FM)-based text-to-speech (TTS) models within a coarse-to-fine generation paradigm. Unlike con…