Zecheng Hao

Zecheng Hao

Auto-generated profile, based on publicly available information

Zecheng Hao

Peking University, Beijing, Beijing, CHINA

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 October 2024
Verified
Conference Paper
Open access
ZHZihan Huang
Zihan Huang
XSXinyu Shi
Xinyu Shi
ZHZecheng Hao
Zecheng Hao
TBTong Bu
Tong Bu
JDJianhao Ding
Jianhao Ding
THTiejun Huang
Tiejun Huang
Spiking neural networks (SNNs) show great potential due to their energy efficiency, fast processing capabilities, and robustness. There are two main approaches to constructing SNNs. Direct training methods require much memory, while conversion methods offer a simpler and more efficient option. However, current conversion methods mainly focus on converting convolutional neural networks (CNNs) to SNNs. Converting Transformers to SNN is challenging because of the presence of non-linear modules. In this paper, we propose an Expectation Compensation Module to preserve the accuracy of the conversion. The core idea is to use information from the previous T time-steps to calculate the expected output at time-step T. We also propose a Multi-Threshold Neuron and the corresponding Parallel Parameter normalization to address the challenge of large time steps needed for high accuracy, aiming to reduce network latency and power consumption. Our experimental results demonstrate that our approach achieves state-of-the-art performance. For example, we achieve a top-1 accuracy of 88.60% with only a 1% loss in accuracy using 4 time steps while consuming only 35% of the original power of the Transformer. To our knowledge, this is the first successful Artificial Neural Network (ANN) to SNN conversion for Spiking Transformers that achieves high accuracy, low latency, and low power consumption on complex datasets. The source codes of the proposed method are available at https://github.com/h-z-h-cell/Transformer-to-SNN-ECMT.

15 July 2025 in Neural Networks
Verified
Journal Article
Closed access
YGYuFei Guo
YuFei Guo
ZHZecheng Hao
Zecheng Hao
JSJiahang Shao
Jiahang Shao
JZJie Zhou
Jie Zhou
XLXiaode Liu
Xiaode Liu
ZMZhe Ma
Zhe Ma

19 December 2024 in IEEE Geoscience and Remote Sensing Letters
Verified
Journal Article
Closed access
ZWZeyu Wang
Zeyu Wang
ZHZecheng Hao
Zecheng Hao
YZYuhan Zhang
Yuhan Zhang
YFYuchao Feng
Yuchao Feng
YGYuFei Guo
YuFei Guo
Remote sensing (RS) technology has become essential for monitoring urban development, including applications like population growth analysis, transportation congestion forecasting, and climate change detection. However, its potential for future urban planning, particularly in predicting future urban layouts remains largely unexplored. This study introduces UP-Diff, a novel method leveraging generative models for urban planning, to address this gap. UP-Diff leverages information from current urban layouts and planned change maps to predict future urban configurations. Key challenges addressed include the integration of urban layouts and change maps into LDM through careful architecture improvements and the mitigation of limited training data by employing a pre-trained Stable Diffusion model with fixed weights, trainable ConvNeXt, and trainable cross-attention layers. Our method significantly streamlines the urban planning process by automating layout predictions, thus reducing the time and effort required compared to traditional manual methods. Comprehensive evaluations on the LEVIR-CD and SYSU-CD datasets validate that UP-Diff achieves high-fidelity predictions of future urban layouts, demonstrating its effectiveness and potential for advancing remote sensing-based urban planning methodologies. Our code and model weights are available at https://github.com/zeyuwang-zju/UP-Diff.

26 July 2026
Verified
Conference Paper
Open access
QMQichao Ma
Qichao Ma
ZHZecheng Hao
Zecheng Hao
ZZZihao Zheng
Zihao Zheng
LLLing Liang
Ling Liang
XHXing Hu
Xing Hu
THTiejun Huang
Tiejun Huang
Online LLM inference service demands low latency with tight memory. Mixture of Experts (MoE) meets memory constraints by activating a few Experts, computing active ones on GPUs, and offloading others to CPUs. For latency, acceleration systems utilize Expert caching and prefetching strategies, assuming Experts' temporal locality and predictable prefetch. When these assumptions fail, wrong experts loads inflate latency, enabling Denial-of-Service (DoS). Existing LLM DoS attacks target extending LLMs' generation length, with high attack costs, weak transferability, and lack MoE analysis. In this work, we expose this vulnerability in GPU-centric MoE offloading acceleration and present Icarus, a gradient-based universal attack injecting an adversarial prefix embedding to disable such acceleration. Icarus first incorporates Cache Temporal Locality Minimization (TLM) and Prefetch Expert Prediction Misleading (EPM) to model MoE decoding behavior systematically, then adversarially breaks acceleration. Next, a scheduler balances attack targets. Experiments show that, by applying a single prefix embedding before any user input, Icarus increases Expert replacements by 85.4%, and decreases prefetch accuracy by 12.5%, achieving an average 0.7X decoding slowdown under SOTA GPU-centric acceleration strategies across models and devices.

9 November 2026
Verified
Conference Paper
Open access
ZHZecheng Hao
Zecheng Hao
SXShenghao Xie
Shenghao Xie
KCKang Chen
Kang Chen
WLWenxuan Liu
Wenxuan Liu
ZYzhaofei yu
zhaofei yu
THTiejun Huang
Tiejun Huang