8 citations · 25 across the 39 of their papers we have counts for
10 papers · 1 filter
Reward Transport: Property Control in Flow Matching via Noise-Space Alignment
Kehan Guo, Yili Shen, Yujun Zhou +4
The coupling in flow matching -- the rule pairing noise vectors with data points -- is typically treated as a computational choice. We show that this coupling can instead serve as…
Getting Better at Working With You: Compiling User Corrections into Runtime Enforcement for Coding Agents
Yujun Zhou, Kehan Guo, Haomin Zhuang +8
Interactive LLM agents are becoming part of daily work, but they do not reliably become easier to work with over time: a correction remembered in one session may still be violated…
DOG-DPO:Dynamic Optimization in Geometry for Safety Alignment
Yi Nian, Tiankai Yang, Yudi Zhang +7
Safety alignment for large language models relies on preference data, but current pipelines often train on large, redundant datasets. Existing data selection methods typically scor…
SkillGen: Verified Inference-Time Agent Skill Synthesis
Yuchen Ma, Yue Huang, Han Bao +5
Skills are a promising way to improve LLM agent capabilities without retraining, while keeping the added procedure reusable and controllable. However, high-quality skills are still…
Why Search When You Can Transfer? Amortized Agentic Workflow Design from Structural Priors
Shiyi Du, Jiayuan Liu, Weihua Du +6
Automated agentic workflow design currently relies on per-task iterative search, which is computationally prohibitive and fails to reuse structural knowledge across tasks. We obser…
Alignment Risks from Capability-Seeking RL Training
Yujun Zhou, Yue Huang, Han Bao +8
While most AI alignment research focuses on preventing models from generating explicitly harmful content, a more subtle risk arises from capability-seeking RL training in vulnerabl…
Building a Foundational Guardrail for General Agentic Systems via Synthetic Data
Yue Huang, Hang Hua, Yujun Zhou +11
While LLM agents can plan multi-step tasks, intervening at the planning stage-before any action is executed-is often the safest way to prevent harm, since certain risks can lead to…
Causally-Enhanced Reinforcement Policy Optimization
Xiangqi Wang, Yue Huang, Yujun Zhou +3
Large language models (LLMs) trained with reinforcement objectives often achieve superficially correct answers via shortcut strategies, pairing correct outputs with spurious or unf…
Beyond Single-Value Metrics: Evaluating and Enhancing LLM Unlearning with Cognitive Diagnosis
Yicheng Lang, Kehan Guo, Yue Huang +5
Due to the widespread use of LLMs and the rising critical ethical and safety concerns, LLM unlearning methods have been developed to remove harmful knowledge and undesirable capabi…
Preference Leakage: A Contamination Problem in LLM-as-a-judge
Dawei Li, Renliang Sun, Yue Huang +6
Large Language Models (LLMs) as judges and LLM-based data synthesis have emerged as two fundamental LLM-driven data annotation methods in model development. While their combination…