Zheng Xu

The Impact of Neural Network Overparameterization on Gradient Confusion and Stochastic Gradient Descent

2019 · arXiv (Cornell University)

This paper studies how neural network architecture affects the speed of training. We introduce a simple concept called gradient confusion to help formally analyze this. When gradient confusion is high, stochastic gradients produced by different …

Abnormal Behavior Recognition System based on Improved CRNN Model

2022 · 2022 IEEE International Conference on Artificial Intelligence and Computer Applications (ICAICA)

Abnormal behavior recognition technology based on surveillance video has been an important direction of current research, with high research value and application demand. Due to the complexity and variability of crowd movement and external environment, …

User Inference Attacks on Large Language Models

2023 · arXiv (Cornell University)

Fine-tuning is a common and effective method for tailoring large language models (LLMs) to specialized tasks and applications. In this paper, we study the privacy implications of fine-tuning LLMs on user data. To this end, …

Enhancing Problem-Solving Abilities with Reinforcement Learning-Augmented Large Language Models

2024

The problem-solving abilities of large language models (LLMs) have significantly advanced in various fields, yet their potential for tackling complex tasks remains underexplored. This paper introduces a novel framework that enhances LLM performance by incorporating …

UpSafe$^\circ$C: Upcycling for Controllable Safety in Large Language Models

2025 · arXiv (Cornell University)

Large Language Models (LLMs) have achieved remarkable progress across a wide range of tasks, but remain vulnerable to safety risks such as harmful content generation and jailbreak attacks. Existing safety techniques -- including external guardrails, …