58 citations · 112 across the 61 of their papers we have counts for
65 papers
Can Agents Keep Secrets They Were Never Told? Benchmarking Latent Privacy Risks Across Sessions with MemJoin
Qishun Yang, Shu Yang, Difei Xu +2
To support sustained interactions and personalized assistance, agents use long-term memory to retain and combine user information across sessions. Reusing such memories can lead an…
LogicTrack: Auditing Reasoning Trajectories of Large Language Models with Formal Logic Solvers
Jingyu Hu, Shu Yang, Weiru Liu +1
Chain-of-Thought (CoT) reasoning has been shown to improve the performance of large language models (LLMs), yet existing optimization methods largely rely on outcome-based feedback…
Neuron-Guided Fine-Tuning: Unlocking Efficient Alignment Mechanisms for Large Language Models
Zeyu Wu, Junchao Wu, Shudong Liu +8
Existing Supervised Fine-Tuning paradigms, particularly Full Parameter Fine-Tuning are often plagued by parameter redundancy, inconsistent data quality, and catastrophic forgetting…
Final Checkpoints Are Not Enough: Analyzing Latent Reasoning Faithfulness Along Training Trajectories
Hengyu Jin, Shu Yang, Di Wang
Latent reasoning performs multi-step inference in continuous hidden states, promising more compact and efficient reasoning. However, these opaque states raise a question of faithfu…
Mitigating Reasoning-Induced Misalignment via Safety-Direction Penalty
Yipeng Zhao, Qishun Yang, Shenzhe Zhu +2
Reasoning-Induced Misalignment, where fine-tuning on reasoning data containing no harmful content, including mathematics, code, and problem-solving with chain-of-thought traces can…
Unanimity Without Persuasion: A Single Round of Debate Erases the Disagreement That Verification Needs
Yang Shu
A debate panel can become unanimous without becoming more correct. This is dangerous for downstream safeguards: a substituted verification ballot can change only narrow-margin vote…
When Verifiers Vote Backwards under Verdict Substitution: Signed Pivotal Value in Correlated Self-Consistency
Yang Shu
Replacing one ballot can change a majority decision only on queries decided by a single vote; this structural fact requires no independence assumption. We study the sign of that ch…
Forged Peer Judgments Mislead Multimodal LLM Judge Panels: Source-Blind Anchoring and Panel-Consensus Verification
Yang Shu
Multimodal LLM judge panels can cross-reference peers, but a quoted peer judgment may itself be untrusted. We expose source-blind anchoring as a text-level attack surface in vision…
Blind to the Pivotal Vote: Aggregate Independence Metrics Miss Where Verification Actually Helps
Yang Shu
LLM judge panels are a standard evaluation tool, but prior work reports highly correlated panel errors: nine judges provide roughly the effective information of two independent one…
AISPA: User-Centric System Prompt Auditing for Large Language Model Applications
Xiangning Lin, Shenzhe Zhu, Shu Yang +23
System prompts are instructions configured by developers to govern the behaviors of foundation models in AI applications. They are used throughout commercial AI products, but are r…
Humanly: A Configurable and Traceable Environment for Human-AI Collaborative Writing
Shenzhe Zhu, Haoqian Zhang, Xu Yang +7
Teachers, conference chairs, and public readers all judge writing from limited evidence, seeing only a finished document and not the process that produced it. Final text alone cann…
ProACT: Towards Breakdown-Aware Proactive Agent in Multi-User Collaboration
Shu Yang, Difei Xu, Jiaxin Pei +1
Conversational agents are increasingly embedded in human collaborative work, yet they remain fundamentally passive and reactive: they respond to explicit user requests rather than…
SelfMem: Self-Optimizing Memory for AI Agents
Shu Yang, Junchao Wu, Derek F. Wong +1
While current AI agents support increasingly long context windows, tool use, and skill execution for long-horizon tasks, they still require memory systems to effectively leverage h…
Multi-User Large Language Model Agents
Shu Yang, Shenzhe Zhu, Hao Zhu +5
Large language models (LLMs) and LLM-based agents are increasingly deployed as assistants in planning and decision making, yet most existing systems are implicitly optimized for a…
Visual Self-Fulfilling Alignment: Shaping Safety-Oriented Personas via Threat-Related Images
Qishun Yang, Shu Yang, Lijie Hu +1
Multimodal large language models (MLLMs) face safety misalignment, where visual inputs enable harmful outputs. To address this, existing methods require explicit safety labels or c…
STARS: Skill-Triggered Audit for Request-Conditioned Invocation Safety in Agent Systems
Guijia Zhang, Shu Yang, Xilin Gong +1
Autonomous language-model agents increasingly rely on installable skills and tools to complete user tasks. Static skill auditing can expose capability surface before deployment, bu…
Hierarchical Alignment: Enforcing Hierarchical Instruction-Following in LLMs through Logical Consistency
Shu Yang, Zihao Zhou, Di Wang +1
Large language models increasingly operate under multiple instructions from heterogeneous sources with different authority levels, including system policies, user requests, tool ou…
Concept-Based Dictionary Learning for Inference-Time Safety in Vision Language Action Models
Siqi Wen, Shu Yang, Shaopeng Fu +3
Vision Language Action (VLA) models close the perception action loop by translating multimodal instructions into executable behaviors, but this very capability magnifies safety ris…
Neuron-Aware Data Selection In Instruction Tuning For Large Language Models
Xin Chen, Junchao Wu, Shu Yang +6
Instruction Tuning (IT) has been proven to be an effective approach to unlock the powerful capabilities of large language models (LLMs). Recent studies indicate that excessive IT d…
Accelerating Suffix Jailbreak attacks with Prefix-Shared KV-cache
Xinhai Wang, Shaopeng Fu, Shu Yang +3
Suffix jailbreak attacks serve as a systematic method for red-teaming Large Language Models (LLMs) but suffer from prohibitive computational costs, as a large number of candidate s…
Word Recovery in Large Language Models Enables Character-Level Tokenization Robustness
Zhipeng Yang, Shu Yang, Lijie Hu +1
Large language models (LLMs) trained with canonical tokenization exhibit surprising robustness to non-canonical inputs such as character-level tokenization, yet the mechanisms unde…
Faithful-Patchscopes: Understanding and Mitigating Model Bias in Hidden Representations Explanation of Large Language Models
Xilin Gong, Shu Yang, Zehua Cao +2
Large Language Models (LLMs) have demonstrated strong capabilities for hidden representation interpretation through Patchscopes, a framework that uses LLMs themselves to generate h…
Hearing is Believing? Evaluating and Analyzing Audio Language Model Sycophancy with SYAUDIO
Junchi Yao, Lokranjan Lakshmikanthan, Annie Zhao +5
Audio Language Models (ALMs) have recently shown strong capabilities in unified reasoning over speech, sound, and natural language; yet we find that they can inherit sycophancy, th…
Not All Code Is Equal: A Data-Centric Study of Code Complexity and LLM Reasoning
Lukas Twist, Shu Yang, Hanqi Yan +4
Large Language Models (LLMs) increasingly exhibit strong reasoning abilities, often attributed to their capacity to generate chain-of-thought-style intermediate reasoning. Recent w…