Xiang Liu

Xiang Liu

Auto-generated profile, based on publicly available information

Xiang Liu

The Hong Kong University of Science and Technology, Hong Kong SAR

Research Focus & Metrics

Academic Career Length

4Years (2023-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
PSPenglei Sun
Penglei Sun
YSYaoxian Song
Yaoxian Song
XLXiang Liu
Xiang Liu
XYXiaofei Yang
Xiaofei Yang
QWQiang Wang
Qiang Wang
XCXiaowen Chu
Xiaowen Chu
3D multimodal question answering (MQA) plays a crucial role in scene understanding by enabling intelligent agents to comprehend their surroundings in 3D environments. While existing research has primarily focused on indoor household tasks and outdoor roadside autonomous driving tasks, there has been limited exploration of city-level scene understanding tasks. Furthermore, existing research faces challenges in understanding city scenes, due to the absence of spatial semantic information and human-environment interaction information at the city level.To address these challenges, we investigate 3D MQA from both dataset and method perspectives. From the dataset perspective, we introduce a novel 3D MQA dataset named City-3DQA for city-level scene understanding, which is the first dataset to incorporate scene semantic and human-environment interactive tasks within the city. From the method perspective, we propose a Scene graph enhanced City-level Understanding method (Sg-CityU), which utilizes the scene graph to introduce the spatial semantic. A new benchmark is reported and our proposed Sg-CityU achieves accuracy of 63.94 % and 63.76 % in different settings of City-3DQA. Compared to indoor 3D MQA methods and zero-shot using advanced large language models (LLMs), Sg-CityU demonstrates state-of-the-art (SOTA) performance in robustness and generalization.

27 October 2025
Verified
Conference Paper
Open access
PSPenglei Sun
Penglei Sun
YSYaoxian Song
Yaoxian Song
XZXiangru Zhu
Xiangru Zhu
XLXiang Liu
Xiang Liu
QWQiang Wang
Qiang Wang
XCXiaowen Chu
Xiaowen Chu

17 November 2023
Verified
Conference Paper
Closed access
YLYi Li
Yi Li
XLXiaobo Liu
Xiaobo Liu
XSXinsheng Song
Xinsheng Song
WZWenjie Zu
Wenjie Zu
XLXiang Liu
Xiang Liu
Inpainting methods based on generative adversarial networks are an important class in the field, and their excellent generative ability can effectively improve the details of the inpainting results. However, generative adversarial networks inherently suffer from training difficulties and model collapse, leading to the large size of models based on such structures and equally difficult training. In addition, most methods incorporate an attention mechanism to improve the inpainting accuracy, which leads to the loss of boundary details of images in the process of restoration, affecting the results. To address the above problems, we propose an inpainting method based on efficient group attention generative adversarial networks (EGA-GAN). The proposed method uses cascaded group attention modules and feature-aligned pyramid networks in the generator. The cascaded group attention module can effectively save computational costs and reduce computational redundancy, and the feature-aligned pyramid network can mitigate the loss of boundary details in the downsampling operation to some extent. Experimental results on two commonly used inpainting datasets, Paris Street View and CelebA-HQ, show that the proposed method can effectively accelerate the speed of restoration and improve the accuracy of results.

29 October 2025 in ACM Transactions on Intelligent Systems and Technology
Verified
Journal Article
Open access
JZJipeng Zhang
Jipeng Zhang
RPRui Pan
Rui Pan
YHYuzheng Hu
Yuzheng Hu
KSKashun Shum
Kashun Shum
GYGuanyu Yao
Guanyu Yao
TZTong Zhang
Tong Zhang
The AI community has witnessed the emergence of various chat-style Large Language Models (LLMs) since the advent of ChatGPT. Despite significant progress in this area, evaluating these models remains a substantial challenge. The evaluations provided by humans or GPT-4 oracles are often taken as the gold standard, but they are neither automatic nor scalable. More recently, a series of (open-source) LLM-based judge models have been introduced, yet they often exhibit model-specific biases, e.g., a LLaMA-family judge favors a LLaMA-family model. On the other hand, autoregressive evaluation metrics, which holds the potential to address the aforementioned issues, remains underexplored. Among them, likelihood-based metrics such as perplexity and negative log-likelihood (NLL) are widely adopted and has proven effective in tracking the pretraining progress of LLMs. However, they struggle to evaluate the generation capabilities of fine-tuned models due to exposure bias , a phenomenon where the distribution of the model’s output gradually deviates from the ground-truth during inference. To address this key issue, in this paper, we propose a novel autoregressive metric, Normalized Discounted Cumulative Gain (NDCG), to improve the evaluation of fine-tuned LLMs. Our experimental results demonstrate that NDCG significantly outperforms likelihood-based metrics: it shows over 45% improvement in both Spearman and Kendall’s tau correlation coefficients for commonsense QA tasks, and aligns more closely with GPT-4 Elo rankings for instruction-tuned models.