Liangwei Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Liangwei Yang has a total of 74 co-authors. Here are the top co-authors by number of shared publications:
- Philip S. Yu
University of Illinois Chicago
21 shared publications - Zhiwei Liu
Salesforce AI Research, Palo Alto, CA, USA
17 shared publications - Mingdai Yang
University of Illinois Chicago, USA
12 shared publications - Chen Wang
University of Illinois at Chicago, Chicago, USA
11 shared publications - Xiaolong Liu
University of Illinois Chicago, Chicago, IL, USA
11 shared publications - Hao Peng
Beihang University, Beijing, China
9 shared publications
Latest Publications
- Open Access
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 5621-5630
GeoGNN: Quantifying and Mitigating Semantic Drift in Text-Attributed Graphs
Graph neural networks (GNNs) on text-attributed graphs (TAGs) typically encode node texts with pretrained language models (PLMs) and propagate the resulting embeddings through linear neighborhood aggregation. Yet PLM representation spaces are geometrically structured, so linear aggregation can distort semantic geometry and cause semantic drift, where aggregated representations move away from the underlying manifold. We introduce a local PCA-based drift metric to quantify this effect across aggregation mechanisms, and propose Geodesic Aggregation, which aggregates neighbors along geodesics with log--exp mappings on the unit sphere, keeping representations on the spherical proxy while retaining unit-norm angular structure. GeoGNN instantiates this mechanism with spherical attention and manifold interpolation. Experiments on four benchmark datasets with multiple text encoders show that GeoGNN mitigates semantic drift and consistently outperforms strong baselines.
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 5621-5630
Open AccessGeoGNN: Quantifying and Mitigating Semantic Drift in Text-Attributed Graphs
Graph neural networks (GNNs) on text-attributed graphs (TAGs) typically encode node texts with pretrained language models (PLMs) and propagate the resulting embeddings through linear neighborhood aggregation. Yet PLM representation spaces are geometrically structured, so linear aggregation can distort semantic geometry and cause semantic drift, where aggregated representations move away from the underlying manifold. We introduce a local PCA-based drift metric to quantify this effect across aggregation mechanisms, and propose Geodesic Aggregation, which aggregates neighbors along geodesics with log--exp mappings on the unit sphere, keeping representations on the spherical proxy while retaining unit-norm angular structure. GeoGNN instantiates this mechanism with spherical attention and manifold interpolation. Experiments on four benchmark datasets with multiple text encoders show that GeoGNN mitigates semantic drift and consistently outperforms strong baselines.
- Open Access
- 20 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1578-1589
StabCF: A Stabilized Training Method for Collaborative Filtering
- 20 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1578-1589
Open AccessStabCF: A Stabilized Training Method for Collaborative Filtering
- Open Access
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 5356-5360
Dual Context-Aware Negative Sampling Strategy for Graph-based Collaborative Filtering
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 5356-5360
Open AccessDual Context-Aware Negative Sampling Strategy for Graph-based Collaborative Filtering
- Open Access
- 22 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 671-676
SGCL: Unifying Self-Supervised and Supervised Learning for Graph Recommendation
- 22 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 671-676
Open AccessSGCL: Unifying Self-Supervised and Supervised Learning for Graph Recommendation
- Open Access
- 3 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 5866-5875
Benchmarking Graph Foundation Models
- 3 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 5866-5875
Open AccessBenchmarking Graph Foundation Models
- Open Access
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 888-892
Graph-Sequential Alignment and Uniformity: Toward Enhanced Recommendation Systems
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 888-892
Open AccessGraph-Sequential Alignment and Uniformity: Toward Enhanced Recommendation Systems
- Open Access
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 1470-1474
Training Large Recommendation Models via Graph-Language Tokens Alignment
- 8 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 1470-1474
Open AccessTraining Large Recommendation Models via Graph-Language Tokens Alignment
- Open Access
RAU: Towards Regularized Alignment and Uniformity for Representation Learning in Recommendation
Recommender systems (RecSys) have become essential in modern society, driving user engagement and satisfaction across diverse online platforms. Most RecSys focuses on designing a powerful encoder to embed users and items into high-dimensional vector representation space, with loss functions optimizing their representation distributions. Recent studies reveal that directly optimizing key properties of the representation distribution, such as alignment and uniformity, can outperform complex encoder designs. However, existing methods for optimizing critical attributes overlook the impact of dataset sparsity on the model: limited user-item interactions lead to sparse alignment, while excessive interactions result in uneven uniformity, both of which degrade performance. In this paper, we identify the sparse alignment and uneven uniformity issues, and further propose Regularized Alignment and Uniformity (RAU) to cope with these two issues accordingly. RAU consists of two novel regularization methods for alignment and uniformity to learn better user/item representation. 1) Center-strengthened alignment further aligns the average in-batch user/item representation to provide an enhanced alignment signal and further minimize the disparity between user and item representation. 2) Low-variance-guided uniformity minimizes the variance of pairwise distances along with uniformity, which provides extra guidance to a more stabilized uniformity increase during training. We conducted extensive experiments on three real-world datasets, and the proposed RAU resulted in significant performance improvements compared to current state-of-the-art CF methods, which confirms the advantages of the two proposed regularization methods.
Open AccessRAU: Towards Regularized Alignment and Uniformity for Representation Learning in Recommendation
Recommender systems (RecSys) have become essential in modern society, driving user engagement and satisfaction across diverse online platforms. Most RecSys focuses on designing a powerful encoder to embed users and items into high-dimensional vector representation space, with loss functions optimizing their representation distributions. Recent studies reveal that directly optimizing key properties of the representation distribution, such as alignment and uniformity, can outperform complex encoder designs. However, existing methods for optimizing critical attributes overlook the impact of dataset sparsity on the model: limited user-item interactions lead to sparse alignment, while excessive interactions result in uneven uniformity, both of which degrade performance. In this paper, we identify the sparse alignment and uneven uniformity issues, and further propose Regularized Alignment and Uniformity (RAU) to cope with these two issues accordingly. RAU consists of two novel regularization methods for alignment and uniformity to learn better user/item representation. 1) Center-strengthened alignment further aligns the average in-batch user/item representation to provide an enhanced alignment signal and further minimize the disparity between user and item representation. 2) Low-variance-guided uniformity minimizes the variance of pairwise distances along with uniformity, which provides extra guidance to a more stabilized uniformity increase during training. We conducted extensive experiments on three real-world datasets, and the proposed RAU resulted in significant performance improvements compared to current state-of-the-art CF methods, which confirms the advantages of the two proposed regularization methods.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 2016-2024
LLMInit: A Free Lunch from Large Language Models for Selective Initialization of Recommendation
Weizhi Zhang, Liangwei Yang, Wooseong Yang, Henry Peng Zou, Yuqing Liu, Ke Xu, Sourav Medya, Philip S. Yu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track. 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 2016-2024
Open AccessLLMInit: A Free Lunch from Large Language Models for Selective Initialization of Recommendation
Weizhi Zhang, Liangwei Yang, Wooseong Yang, Henry Peng Zou, Yuqing Liu, Ke Xu, Sourav Medya, Philip S. Yu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track. 2025.
- Open Access
- 21 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 2315-2325
Collaborative Alignment for Recommendation
Traditional recommender systems have primarily relied on identity representations (IDs) to model users and items. Recently, the integration of pre-trained language models (PLMs) has enhanced the capability to capture semantic descriptions of items. However, while PLMs excel in few-shot, zero-shot, and unified modeling scenarios, they often overlook the crucial signals from collaborative filtering (CF), resulting in suboptimal performance when sufficient training data is available. To effectively combine semantic representations with the CF signal and enhance recommender system performance in both warm and cold settings, two major challenges must be addressed: (1) bridging the gap between semantic and collaborative representation spaces, and (2) refining while preserving the integrity of semantic representations. In this paper, we introduce CARec, a novel model that adeptly integrates collaborative filtering signals with semantic representations, ensuring alignment within the semantic space while maintaining essential semantics. We present experimental results from four real-world datasets, which demonstrate significant improvements. By leveraging collaborative alignment, CARec also shows remarkable effectiveness in cold-start scenarios, achieving notable enhancements in recommendation performance. The code is available at https://github.com/ChenMetanoia/CARec **REMOVE 2nd URL**://github.com/ChenMetanoia/CARec.
- 21 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 2315-2325
Open AccessCollaborative Alignment for Recommendation
Traditional recommender systems have primarily relied on identity representations (IDs) to model users and items. Recently, the integration of pre-trained language models (PLMs) has enhanced the capability to capture semantic descriptions of items. However, while PLMs excel in few-shot, zero-shot, and unified modeling scenarios, they often overlook the crucial signals from collaborative filtering (CF), resulting in suboptimal performance when sufficient training data is available. To effectively combine semantic representations with the CF signal and enhance recommender system performance in both warm and cold settings, two major challenges must be addressed: (1) bridging the gap between semantic and collaborative representation spaces, and (2) refining while preserving the integrity of semantic representations. In this paper, we introduce CARec, a novel model that adeptly integrates collaborative filtering signals with semantic representations, ensuring alignment within the semantic space while maintaining essential semantics. We present experimental results from four real-world datasets, which demonstrate significant improvements. By leveraging collaborative alignment, CARec also shows remarkable effectiveness in cold-start scenarios, achieving notable enhancements in recommendation performance. The code is available at https://github.com/ChenMetanoia/CARec **REMOVE 2nd URL**://github.com/ChenMetanoia/CARec.