Liangcheng Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Liangcheng Li has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Jiajun Bu
Zhejiang Key Lab of Accessible Perception & Intelligent Systems, Zhejiang University, Hangzhou, China
9 shared publications - Sheng Zhou
Zhejiang Key Lab of Accessible Perception & Intelligent Systems, Zhejiang University, Hangzhou, China
7 shared publications - Zhi Yu
Zhejiang University
4 shared publications - Gang Huang
School of Software Technology, Zhejiang University, Ningbo, Zhejiang, China
2 shared publications - Haishuai Wang
Zhejiang University
2 shared publications - Jingru Li
College of Computer Science and Technology, Zhejiang University, Zheda Rd., Hangzhou, 310027, Zhejiang, China
2 shared publications
Latest Publications
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 102-112
ChartAccessMobile: An Intelligent System for Accessible Chart Navigation on Mobile Applications
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 102-112
Open AccessChartAccessMobile: An Intelligent System for Accessible Chart Navigation on Mobile Applications
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 90-101
Bridging the Semantic Void: A Dual-branch RAG Framework to Enhance GUI Component Description for BVI Users
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 90-101
Open AccessBridging the Semantic Void: A Dual-branch RAG Framework to Enhance GUI Component Description for BVI Users
- Open Access
Causal-LLaVA: Causal Disentanglement for Mitigating Hallucination in Multimodal Large Language Models
Multimodal Large Language Models (MLLMs) have demonstrated strong performance in visual understanding tasks, yet they often suffer from object hallucinations--generating descriptions of objects that are inconsistent with or entirely absent from the input. This issue is closely related to dataset biases, where frequent co-occurrences of objects lead to entangled semantic representations across modalities. As a result, models may erroneously activate object representations that are commonly associated with the input but not actually present. To address this, we propose a causality-driven disentanglement framework that mitigates hallucinations through causal intervention. Our approach includes a Causal-Driven Projector in the visual pathway and a Causal Intervention Module integrated into the final transformer layer of the language model. These components work together to reduce spurious correlations caused by biased training data. Experimental results show that our method significantly reduces hallucinations while maintaining strong performance on multiple multimodal benchmarks. Visualization analyses further confirm improved separability of object representations. The code is available at: https://github.com/IgniSavium/Causal-LLaVA
Open AccessCausal-LLaVA: Causal Disentanglement for Mitigating Hallucination in Multimodal Large Language Models
Multimodal Large Language Models (MLLMs) have demonstrated strong performance in visual understanding tasks, yet they often suffer from object hallucinations--generating descriptions of objects that are inconsistent with or entirely absent from the input. This issue is closely related to dataset biases, where frequent co-occurrences of objects lead to entangled semantic representations across modalities. As a result, models may erroneously activate object representations that are commonly associated with the input but not actually present. To address this, we propose a causality-driven disentanglement framework that mitigates hallucinations through causal intervention. Our approach includes a Causal-Driven Projector in the visual pathway and a Causal Intervention Module integrated into the final transformer layer of the language model. These components work together to reduce spurious correlations caused by biased training data. Experimental results show that our method significantly reduces hallucinations while maintaining strong performance on multiple multimodal benchmarks. Visualization analyses further confirm improved separability of object representations. The code is available at: https://github.com/IgniSavium/Causal-LLaVA
- Open Access
- 25 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-17
TableNarrator: Making Image Tables Accessible to Blind and Low Vision People
- 25 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-17
Open AccessTableNarrator: Making Image Tables Accessible to Blind and Low Vision People
- Open Access
- 13 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 160-164
Making Accessible Movies Easily: An Intelligent Tool for Authoring and Integrating Audio Descriptions to Movies
- 13 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 160-164
Open AccessMaking Accessible Movies Easily: An Intelligent Tool for Authoring and Integrating Audio Descriptions to Movies
- 10 May 2024
- Published by Elsevier in Neural Networks
- Vol. 177, 106386
AdaDFKD: Exploring adaptive inter-sample relationship in data-free knowledge distillation
- 10 May 2024
- Published by Elsevier in Neural Networks
- Vol. 177, 106386
AdaDFKD: Exploring adaptive inter-sample relationship in data-free knowledge distillation
- Open Access
Dynamic Data-Free Knowledge Distillation by Easy-to-Hard Learning Strategy
Data-free knowledge distillation (DFKD) is a widely-used strategy for Knowledge Distillation (KD) whose training data is not available. It trains a lightweight student model with the aid of a large pretrained teacher model without any access to training data. However, existing DFKD methods suffer from inadequate and unstable training process, as they do not adjust the generation target dynamically based on the status of the student model during learning. To address this limitation, we propose a novel DFKD method called CuDFKD. It teaches students by a dynamic strategy that gradually generates easy-to-hard pseudo samples, mirroring how humans learn. Besides, CuDFKD adapts the generation target dynamically according to the status of student model. Moreover, We provide a theoretical analysis of the majorization minimization (MM) algorithm and explain the convergence of CuDFKD. To measure the robustness and fidelity of DFKD methods, we propose two more metrics, and experiments shows CuDFKD has comparable performance to state-of-the-art (SOTA) DFKD methods on all datasets. Experiments also present that our CuDFKD has the fastest convergence and best robustness over other SOTA DFKD methods.
Open AccessDynamic Data-Free Knowledge Distillation by Easy-to-Hard Learning Strategy
Data-free knowledge distillation (DFKD) is a widely-used strategy for Knowledge Distillation (KD) whose training data is not available. It trains a lightweight student model with the aid of a large pretrained teacher model without any access to training data. However, existing DFKD methods suffer from inadequate and unstable training process, as they do not adjust the generation target dynamically based on the status of the student model during learning. To address this limitation, we propose a novel DFKD method called CuDFKD. It teaches students by a dynamic strategy that gradually generates easy-to-hard pseudo samples, mirroring how humans learn. Besides, CuDFKD adapts the generation target dynamically according to the status of student model. Moreover, We provide a theoretical analysis of the majorization minimization (MM) algorithm and explain the convergence of CuDFKD. To measure the robustness and fidelity of DFKD methods, we propose two more metrics, and experiments shows CuDFKD has comparable performance to state-of-the-art (SOTA) DFKD methods on all datasets. Experiments also present that our CuDFKD has the fastest convergence and best robustness over other SOTA DFKD methods.
- 24 May 2023
- Published by Elsevier in Information Sciences
- Vol. 642
Dynamic data-free knowledge distillation by easy-to-hard learning strategy
- 24 May 2023
- Published by Elsevier in Information Sciences
- Vol. 642
Dynamic data-free knowledge distillation by easy-to-hard learning strategy
- 26 June 2018
- Published by Springer Nature
- p. 559-562
China’s Efforts to Information Service for Visual-Impaired People
- 26 June 2018
- Published by Springer Nature
- p. 559-562
China’s Efforts to Information Service for Visual-Impaired People
- 2 October 2017
- Published by Emerald Publishing in Internet Research
- Vol. 27 (5) , 1190-1208
An optimal sampling method for web accessibility quantitative metric and its online extension
As existing studies show the accuracy of sampling methods depends heavily on the evaluation metric in web accessibility evaluation, the purpose of this paper is to propose a sampling method OPS-WAQM optimized for Web Accessibility Quantitative Metric (WAQM). Furthermore, to support quick accessibility evaluation or real-time website accessibility monitoring, the authors also provide online extension for the sampling method. In the OPS-WAQM method, the authors propose a minimal sampling error model for WAQM and use a greedy algorithm to approximately solve the optimization problem to determine the sample numbers in different layers. To make OPS-WAQM online, the authors apply the sampling in crawling strategy. The sampling method OPS-WAQM and its online extension can both achieve good sampling quality by choosing the optimal sample numbers in different layers. Moreover, the online extension can also support quick accessibility evaluation by sampling and evaluating the pages in crawling. To the best of the authors’ knowledge, the sampling method OPS-WAQM in this paper is the first attempt to optimize for a specific evaluation metric. Meanwhile, the online extension not only greatly reduces the serious I/O issues in existing web accessibility evaluation, but also supports quick web accessibility evaluation by sampling in crawling.
- 2 October 2017
- Published by Emerald Publishing in Internet Research
- Vol. 27 (5) , 1190-1208
An optimal sampling method for web accessibility quantitative metric and its online extension
As existing studies show the accuracy of sampling methods depends heavily on the evaluation metric in web accessibility evaluation, the purpose of this paper is to propose a sampling method OPS-WAQM optimized for Web Accessibility Quantitative Metric (WAQM). Furthermore, to support quick accessibility evaluation or real-time website accessibility monitoring, the authors also provide online extension for the sampling method. In the OPS-WAQM method, the authors propose a minimal sampling error model for WAQM and use a greedy algorithm to approximately solve the optimization problem to determine the sample numbers in different layers. To make OPS-WAQM online, the authors apply the sampling in crawling strategy. The sampling method OPS-WAQM and its online extension can both achieve good sampling quality by choosing the optimal sample numbers in different layers. Moreover, the online extension can also support quick accessibility evaluation by sampling and evaluating the pages in crawling. To the best of the authors’ knowledge, the sampling method OPS-WAQM in this paper is the first attempt to optimize for a specific evaluation metric. Meanwhile, the online extension not only greatly reduces the serious I/O issues in existing web accessibility evaluation, but also supports quick web accessibility evaluation by sampling in crawling.