Yu Shang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Yu Shang has a total of 62 co-authors. Here are the top co-authors by number of shared publications:
- Yong Li
Tsinghua University
11 shared publications - Chen Gao
Tsinghua University
5 shared publications - Depeng Jin
Beijing National Research Center for Information Science and Technology (BNRist), Department of Electronic Engineering, Tsinghua University, Beijing, China
4 shared publications - Jiansheng Chen
University of Science and Technology Beijing, Beijing, China
4 shared publications - Fengli Xu
Tsinghua University
2 shared publications - Jie Feng
Tsinghua University, China
2 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 10528-10537
UrbanWorld2.0: A Multimodal Agentic Framework for Reality-Aligned 3D World Generation at City-Scale
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 10528-10537
Open AccessUrbanWorld2.0: A Multimodal Agentic Framework for Reality-Aligned 3D World Generation at City-Scale
- Open Access
- 1 January 2026
- Published by KEAI Publishing in AI Open
- Vol. 7, 208-226
WorldArena: A unified benchmark for evaluating perception and functional utility of embodied world models
- 1 January 2026
- Published by KEAI Publishing in AI Open
- Vol. 7, 208-226
Open AccessWorldArena: A unified benchmark for evaluating perception and functional utility of embodied world models
- Open Access
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 12519-12528
AirScape: An Aerial Generative World Model with Motion Controllability
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 12519-12528
Open AccessAirScape: An Aerial Generative World Model with Motion Controllability
- Top Cited PapersOpen Access
- 9 September 2025
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 58 (3) , 1-38
Understanding World or Predicting Future? A Comprehensive Survey of World Models
The concept of world models has garnered significant attention due to advancements in multimodal large language models such as GPT-4 and video generation models such as Sora, which are central to the pursuit of artificial general intelligence. This survey offers a comprehensive review of the literature on world models. Generally, world models are regarded as tools for either understanding the present state of the world or predicting its future dynamics. This review presents a systematic categorization of world models, emphasizing two primary functions: (1) constructing internal representations to understand the mechanisms of the world, and (2) predicting future states to simulate and guide decision-making. Initially, we examine the current progress in these two categories. We then explore the application of world models in key domains, including autonomous driving, robotics, and social simulacra, with a focus on how each domain utilizes these aspects. Finally, we outline key challenges and provide insights into potential future research directions. We summarize the representative articles along with their code repositories in https://github.com/tsinghua-fib-lab/World-Model .
- 9 September 2025
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 58 (3) , 1-38
Top Cited PapersOpen AccessUnderstanding World or Predicting Future? A Comprehensive Survey of World Models
The concept of world models has garnered significant attention due to advancements in multimodal large language models such as GPT-4 and video generation models such as Sora, which are central to the pursuit of artificial general intelligence. This survey offers a comprehensive review of the literature on world models. Generally, world models are regarded as tools for either understanding the present state of the world or predicting its future dynamics. This review presents a systematic categorization of world models, emphasizing two primary functions: (1) constructing internal representations to understand the mechanisms of the world, and (2) predicting future states to simulate and guide decision-making. Initially, we examine the current progress in these two categories. We then explore the application of world models in key domains, including autonomous driving, robotics, and social simulacra, with a focus on how each domain utilizes these aspects. Finally, we outline key challenges and provide insights into potential future research directions. We summarize the representative articles along with their code repositories in https://github.com/tsinghua-fib-lab/World-Model .
- Open Access
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 2963-2967
AgentSociety Challenge: Designing LLM Agents for User Modeling and Recommendation on Web Platforms
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 2963-2967
Open AccessAgentSociety Challenge: Designing LLM Agents for User Modeling and Recommendation on Web Platforms
- Open Access
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 793-796
A Large-scale Dataset with Behavior, Attributes, and Content of Mobile Short-video Platform
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 793-796
Open AccessA Large-scale Dataset with Behavior, Attributes, and Content of Mobile Short-video Platform
- Open Access
- 13 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 4697-4705
Improving Item-side Fairness of Multimodal Recommendation via Modality Debiasing
Multimodal recommender systems have acquired applications in broad web scenarios such as e-commerce businesses and short-video platforms. Existing multimodal recommendation methods generally boost performance by introducing item-side multimodal content as supplement information. However, the common training paradigm, i.e., encoding unimodal content respectively and fusing them to fit user preference scores, makes the model biased towards items with prevailing modality content under non-uniform training data. This results in a serious item-side unfairness issue, i.e., some items with prevailing modality content are over-recommended while a large number of items don't receive adequate recommendation opportunities, leaving corresponding content providers at great disadvantage. Aiming to eliminate such modality bias and promote item-side fairness, we propose a fairness-aware modality debiasing framework based on counterfactual inference. In the training stage, we additionally introduce unimodal prediction branches to capture the modality bias. In the inference stage, we conduct a fairness-aware counterfactual inference to adaptively eliminate the modality bias. The proposed framework is model-agnostic and flexible to be implemented in various multimodal recommendation models. Extensive experiments on two datasets demonstrate that the proposed method can significantly enhance item-side fairness while providing competitive recommendation accuracy. Our proposed framework is expected to help mitigate the unfair treatment experienced by vulnerable content providers on multimedia web platforms. Codes are available in https://github.com/tsinghua-fib-lab-WWW2024-Modality-Debiasing.
- 13 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 4697-4705
Open AccessImproving Item-side Fairness of Multimodal Recommendation via Modality Debiasing
Multimodal recommender systems have acquired applications in broad web scenarios such as e-commerce businesses and short-video platforms. Existing multimodal recommendation methods generally boost performance by introducing item-side multimodal content as supplement information. However, the common training paradigm, i.e., encoding unimodal content respectively and fusing them to fit user preference scores, makes the model biased towards items with prevailing modality content under non-uniform training data. This results in a serious item-side unfairness issue, i.e., some items with prevailing modality content are over-recommended while a large number of items don't receive adequate recommendation opportunities, leaving corresponding content providers at great disadvantage. Aiming to eliminate such modality bias and promote item-side fairness, we propose a fairness-aware modality debiasing framework based on counterfactual inference. In the training stage, we additionally introduce unimodal prediction branches to capture the modality bias. In the inference stage, we conduct a fairness-aware counterfactual inference to adaptively eliminate the modality bias. The proposed framework is model-agnostic and flexible to be implemented in various multimodal recommendation models. Extensive experiments on two datasets demonstrate that the proposed method can significantly enhance item-side fairness while providing competitive recommendation accuracy. Our proposed framework is expected to help mitigate the unfair treatment experienced by vulnerable content providers on multimedia web platforms. Codes are available in https://github.com/tsinghua-fib-lab-WWW2024-Modality-Debiasing.
- Open Access
- 13 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 1142-1145
Health CLIP: Depression Rate Prediction Using Health Related Features in Satellite and Street View Images
Mental health is a state of mental well-being that enables people to cope with the stresses of life, realize their abilities, learn well and work well, and contribute to their community. It has intrinsic and instrumental value and is integral to our well-being, and its correlation with environmental factors has been a subject of growing interest. As the pressure of society keeps growing, depression has become a severe problem in modern cities, and finding a way to estimate depression rate is of significance to relieve the problem. In this study, we introduce a Contrastive Language-Image Pretraining (CLIP) based novel approach to predict mental health indicators, especially depression rate, through satellite and street view images. Our methodology uses state-of-the-art Multimodal Large Language Model (MLLM), GPT4-vision, to generate health related captions for satellite and street view images, then we use the generated image-text pairs to fine-tune the CLIP model, making its image encoder extract health related features such as green spaces, sports fields, and infrastructral characteristics. The fine-tuning process is employed to bridge the semantic gap between textual descriptions and visual representations, enabling a comprehensive analysis of geo-tagged images. Consequently, our methodology achieves a notable R2 value of 0.565 on prediction of depression rate in New York City with the combination of satellite and street view images. The successful deployment of Health CLIP in a real-world scenario underscores the practical applicability of our approach.
- 13 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 1142-1145
Open AccessHealth CLIP: Depression Rate Prediction Using Health Related Features in Satellite and Street View Images
Mental health is a state of mental well-being that enables people to cope with the stresses of life, realize their abilities, learn well and work well, and contribute to their community. It has intrinsic and instrumental value and is integral to our well-being, and its correlation with environmental factors has been a subject of growing interest. As the pressure of society keeps growing, depression has become a severe problem in modern cities, and finding a way to estimate depression rate is of significance to relieve the problem. In this study, we introduce a Contrastive Language-Image Pretraining (CLIP) based novel approach to predict mental health indicators, especially depression rate, through satellite and street view images. Our methodology uses state-of-the-art Multimodal Large Language Model (MLLM), GPT4-vision, to generate health related captions for satellite and street view images, then we use the generated image-text pairs to fine-tune the CLIP model, making its image encoder extract health related features such as green spaces, sports fields, and infrastructral characteristics. The fine-tuning process is employed to bridge the semantic gap between textual descriptions and visual representations, enabling a comprehensive analysis of geo-tagged images. Consequently, our methodology achieves a notable R2 value of 0.565 on prediction of depression rate in New York City with the combination of satellite and street view images. The successful deployment of Health CLIP in a real-world scenario underscores the practical applicability of our approach.
- Open Access
- 26 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 6274-6282
Enhancing Adversarial Robustness of Multi-modal Recommendation via Modality Balancing
Recently multi-modal recommender systems have been widely applied in real scenarios such as e-commerce businesses. Existing multi-modal recommendation methods exploit the multi-modal content of items as auxiliary information and fuse them to boost performance. Despite the superior performance achieved by multi-modal recommendation models, there's currently no understanding of their robustness to adversarial attacks. In this work, we first identify the vulnerability of existing multi-modal recommendation models. Next, we show the key reason for such vulnerability is modality imbalance, i.e., the prediction score margin between positive and negative samples in the sensitive modality will drop dramatically facing adversarial attacks and fail to be compensated by other modalities. Finally, based on this finding we propose a novel defense method to enhance the robustness of multi-modal recommendation models through modality balancing. Specifically, we first adopt an embedding distillation to obtain a pair of content-similar but prediction-different item embeddings in the sensitive modality and calculate the score margin reflecting the modality vulnerability. Then we optimize the model to utilize the score margin between positive and negative samples in other modalities to compensate for the vulnerability. The proposed method can serve as a plug-and-play module and is flexible to be applied to a wide range of multi-modal recommendation models. Extensive experiments on two real-world datasets demonstrate that our method significantly improves the robustness of multi-modal recommendation models with nearly no performance degradation on clean data.
- 26 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 6274-6282
Open AccessEnhancing Adversarial Robustness of Multi-modal Recommendation via Modality Balancing
Recently multi-modal recommender systems have been widely applied in real scenarios such as e-commerce businesses. Existing multi-modal recommendation methods exploit the multi-modal content of items as auxiliary information and fuse them to boost performance. Despite the superior performance achieved by multi-modal recommendation models, there's currently no understanding of their robustness to adversarial attacks. In this work, we first identify the vulnerability of existing multi-modal recommendation models. Next, we show the key reason for such vulnerability is modality imbalance, i.e., the prediction score margin between positive and negative samples in the sensitive modality will drop dramatically facing adversarial attacks and fail to be compensated by other modalities. Finally, based on this finding we propose a novel defense method to enhance the robustness of multi-modal recommendation models through modality balancing. Specifically, we first adopt an embedding distillation to obtain a pair of content-similar but prediction-different item embeddings in the sensitive modality and calculate the score margin reflecting the modality vulnerability. Then we optimize the model to utilize the score margin between positive and negative samples in other modalities to compensate for the vulnerability. The proposed method can serve as a plug-and-play module and is flexible to be applied to a wide range of multi-modal recommendation models. Extensive experiments on two real-world datasets demonstrate that our method significantly improves the robustness of multi-modal recommendation models with nearly no performance degradation on clean data.
- Open Access
- 21 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 2188-2197
Transferable Structure-based Adversarial Attack of Heterogeneous Graph Neural Network
- 21 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 2188-2197
Open AccessTransferable Structure-based Adversarial Attack of Heterogeneous Graph Neural Network