Qingcai Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Qingcai Chen has a total of 208 co-authors. Here are the top co-authors by number of shared publications:
- Buzhou Tang
School of Computer Science and Technology, Harbin Institute of Technology, Shenzhen, China
35 shared publications - Xiaolong Wang
Harbin Institute of Technology (Shenzhen) Shenzhen China
31 shared publications - Xiangping Wu
College of Information Engineering, China Jiliang University, Hangzhou, China
16 shared publications - Baotian Hu
Harbin Institute of Technology (Shenzhen) Shenzhen China
15 shared publications - Jun Yan
AI Lab, Yidu Cloud (Beijing) Technology Co. Ltd., Beijing, China
11 shared publications - Yang Xiang
Pengcheng Laboratory, Shenzhen, China
11 shared publications
Latest Publications
- Open Access
- 18 November 2026
- Published by MDPI AG in Mathematics
- Vol. 14 (22) , 4243
Large Language Models for Multimodal Fact-Checking: Methodological Advances and Persistent Challenges
Fact-checking aims to assign truth values to claims within a specified context. Unimodal fact-checking methods, due to modality fragmentation and inadequate adaptiveness, struggle in complex scenarios where multimodal misinformation is interconnected. The rapid advancement of generative large language models (LLMs) has provided a novel technological paradigm for multimodal fact-checking. With their capabilities in deep semantic understanding and cross-modal alignment, LLMs can significantly enhance the accuracy, interpretability, and generalizability of misinformation detection. Synthesizing representative studies from 2018 to 2026 retrieved via a multi-dimensional keyword search, two-step screening, and citation chaining, we comprehensively survey the subtask arrangements, LLM-based methodologies, and relevant datasets for both unimodal and multimodal fact-checking tasks. Following the standard fact-checking pipeline, we comprehensively review existing methods across four key stages, namely Claim Detection and Extraction, Evidence Retrieval, Verdict Prediction, and Justification Production, providing a structured perspective for method comparison and evaluation. Our synthesis indicates that while multimodal large language models (MLLMs) enhance cross-modal understanding and verification flexibility, fine-grained visual–textual consistency and evidential sufficiency remain critical challenges. We provide a comprehensive survey of fact-checking research and offer insights into the development of next-generation fact-checking systems expected to be more robust, scalable, and adaptable across diverse information environments.
- 18 November 2026
- Published by MDPI AG in Mathematics
- Vol. 14 (22) , 4243
Open AccessLarge Language Models for Multimodal Fact-Checking: Methodological Advances and Persistent Challenges
Fact-checking aims to assign truth values to claims within a specified context. Unimodal fact-checking methods, due to modality fragmentation and inadequate adaptiveness, struggle in complex scenarios where multimodal misinformation is interconnected. The rapid advancement of generative large language models (LLMs) has provided a novel technological paradigm for multimodal fact-checking. With their capabilities in deep semantic understanding and cross-modal alignment, LLMs can significantly enhance the accuracy, interpretability, and generalizability of misinformation detection. Synthesizing representative studies from 2018 to 2026 retrieved via a multi-dimensional keyword search, two-step screening, and citation chaining, we comprehensively survey the subtask arrangements, LLM-based methodologies, and relevant datasets for both unimodal and multimodal fact-checking tasks. Following the standard fact-checking pipeline, we comprehensively review existing methods across four key stages, namely Claim Detection and Extraction, Evidence Retrieval, Verdict Prediction, and Justification Production, providing a structured perspective for method comparison and evaluation. Our synthesis indicates that while multimodal large language models (MLLMs) enhance cross-modal understanding and verification flexibility, fine-grained visual–textual consistency and evidential sufficiency remain critical challenges. We provide a comprehensive survey of fact-checking research and offer insights into the development of next-generation fact-checking systems expected to be more robust, scalable, and adaptable across diverse information environments.
- Open Access
- 27 April 2026
- Published by Elsevier in SSRN Electronic Journal
Large Language Models for Multimodal Fact-Checking: Methodological Advances and Persistent Challenges
Fact-checking aims to assign truth values to claims within a specified context. Unimodal fact-checking methods, due to modality fragmentation and inadequate adaptiveness, struggle in complex scenarios where multimodal misinformation is interconnected. The recent rapid advancement of generative large language models (LLMs) has provided a novel technological paradigm for multimodal fact-checking. With their capabilities in deep semantic understanding and cross-modal alignment, LLMs can significantly enhance the accuracy, interpretability, and generalizability of misinformation detection. This paper systematically surveys the subtask arrangements, LLM-based methodologies, and relevant datasets for both unimodal and multimodal fact-checking tasks.Following the standard fact-checking pipeline, this paper systematically reviews existing methods from four key stages: Claim Detection and Extraction, Evidence Retrieval, Veracity Prediction, and Justification Production, providing a structured perspective for method comparison and evaluation.This paper provides a systematic survey of fact-checking research and offers insights into the development of next-generation fact-checking systems that are expected to be more robust, scalable, and adaptable to diverse information.
- 27 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessLarge Language Models for Multimodal Fact-Checking: Methodological Advances and Persistent Challenges
Fact-checking aims to assign truth values to claims within a specified context. Unimodal fact-checking methods, due to modality fragmentation and inadequate adaptiveness, struggle in complex scenarios where multimodal misinformation is interconnected. The recent rapid advancement of generative large language models (LLMs) has provided a novel technological paradigm for multimodal fact-checking. With their capabilities in deep semantic understanding and cross-modal alignment, LLMs can significantly enhance the accuracy, interpretability, and generalizability of misinformation detection. This paper systematically surveys the subtask arrangements, LLM-based methodologies, and relevant datasets for both unimodal and multimodal fact-checking tasks.Following the standard fact-checking pipeline, this paper systematically reviews existing methods from four key stages: Claim Detection and Extraction, Evidence Retrieval, Veracity Prediction, and Justification Production, providing a structured perspective for method comparison and evaluation.This paper provides a systematic survey of fact-checking research and offers insights into the development of next-generation fact-checking systems that are expected to be more robust, scalable, and adaptable to diverse information.
- Open Access
WaterSearch: A Quality-Aware Search-based Watermarking Framework for Large Language Models
Watermarking acts as a critical safeguard in text generated by Large Language Models (LLMs). By embedding identifiable signals into model outputs, watermarking enables reliable attribution and enhances the security of machine-generated content. Existing approaches typically embed signals by manipulating token generation probabilities. Despite their effectiveness, these methods inherently face a trade-off between detectability and text quality: the signal strength and randomness required for robust watermarking tend to degrade the performance of downstream tasks. In this paper, we design a novel embedding scheme that controls seed pools to facilitate diverse parallel generation of watermarked text. Based on that scheme, we propose WaterSearch, a sentence-level, search-based watermarking framework adaptable to a wide range of existing methods. WaterSearch enhances text quality by jointly optimizing two key aspects: 1) distribution fidelity and 2) watermark signal characteristics. Furthermore, WaterSearch is complemented by a sentence-level detection method with strong attack robustness. We evaluate our method on three popular LLMs across ten diverse tasks. Extensive experiments demonstrate that our method achieves an average performance improvement of 51.01\% over state-of-the-art baselines at a watermark detectability strength of 95\%. In challenging scenarios such as short text generation and low-entropy output generation, our method yields performance gains of 47.78\% and 36.47\%, respectively. Moreover, under different attack senarios including insertion, synonym substitution and paraphrase attasks, WaterSearch maintains high detectability, further validating its robust anti-attack capabilities. Our code is available at \href{https://github.com/Yukang-Lin/WaterSearch}{https://github.com/Yukang-Lin/WaterSearch}.
Open AccessWaterSearch: A Quality-Aware Search-based Watermarking Framework for Large Language Models
Watermarking acts as a critical safeguard in text generated by Large Language Models (LLMs). By embedding identifiable signals into model outputs, watermarking enables reliable attribution and enhances the security of machine-generated content. Existing approaches typically embed signals by manipulating token generation probabilities. Despite their effectiveness, these methods inherently face a trade-off between detectability and text quality: the signal strength and randomness required for robust watermarking tend to degrade the performance of downstream tasks. In this paper, we design a novel embedding scheme that controls seed pools to facilitate diverse parallel generation of watermarked text. Based on that scheme, we propose WaterSearch, a sentence-level, search-based watermarking framework adaptable to a wide range of existing methods. WaterSearch enhances text quality by jointly optimizing two key aspects: 1) distribution fidelity and 2) watermark signal characteristics. Furthermore, WaterSearch is complemented by a sentence-level detection method with strong attack robustness. We evaluate our method on three popular LLMs across ten diverse tasks. Extensive experiments demonstrate that our method achieves an average performance improvement of 51.01\% over state-of-the-art baselines at a watermark detectability strength of 95\%. In challenging scenarios such as short text generation and low-entropy output generation, our method yields performance gains of 47.78\% and 36.47\%, respectively. Moreover, under different attack senarios including insertion, synonym substitution and paraphrase attasks, WaterSearch maintains high detectability, further validating its robust anti-attack capabilities. Our code is available at \href{https://github.com/Yukang-Lin/WaterSearch}{https://github.com/Yukang-Lin/WaterSearch}.
- 26 November 2025
- Published by Elsevier in Applied Soft Computing
- Vol. 187
Compression of fully-connected layers based on parameter sharing for large-scale category image classification
- 26 November 2025
- Published by Elsevier in Applied Soft Computing
- Vol. 187
Compression of fully-connected layers based on parameter sharing for large-scale category image classification
- 23 November 2025
- Published by Springer Nature
- p. 364-376
CLSurCoder: LLMs Based Cross-Lingual Transfer Learning for Low-Resource Language Surgical Records Coding
- 23 November 2025
- Published by Springer Nature
- p. 364-376
CLSurCoder: LLMs Based Cross-Lingual Transfer Learning for Low-Resource Language Surgical Records Coding
- 30 October 2025
- Published by Springer Nature in Health Information Science and Systems
- Vol. 13 (1) , 1-12
Observe, align, and enhance: a hierarchical retrieval-augmented vision-language model for generating radiology reports
Radiology Report Generation (RRG) is designed to automatically generate diagnostic narratives based on radiological image interpretation, supporting clinicians in making diagnoses and relieving radiologists of reporting pressure. Previous approaches mainly use generative architectures developed to change visibly produced features into coherent written results. However, these approaches generally run into problems with precisely aligning written outputs to visible data, especially in the process of producing thorough diagnostic narratives. We propose a new, hierarchical retrieval-enhanced framework called Observe, Align, and Enhancement (OAE) for the creation of the radiology report to get around these problems: Observe, which leverages retrieval techniques to enhance visual feature comprehension by identifying similar images and associated reports; Align, where the retrieved contextual reports guide the generation of an initial diagnostic report ensuring semantic consistency; and Enhance, an iterative refinement process that incorporates additional textual information to improve semantic coherence and diagnostic precision. The proposed framework is shown in a comprehensive assessment of the IU-XRAY and MIMIC-CXR datasets, which shows that it performs better than the current state-of-the-art approaches, including improved diagnostic accuracy and reported quality.
- 30 October 2025
- Published by Springer Nature in Health Information Science and Systems
- Vol. 13 (1) , 1-12
Observe, align, and enhance: a hierarchical retrieval-augmented vision-language model for generating radiology reports
Radiology Report Generation (RRG) is designed to automatically generate diagnostic narratives based on radiological image interpretation, supporting clinicians in making diagnoses and relieving radiologists of reporting pressure. Previous approaches mainly use generative architectures developed to change visibly produced features into coherent written results. However, these approaches generally run into problems with precisely aligning written outputs to visible data, especially in the process of producing thorough diagnostic narratives. We propose a new, hierarchical retrieval-enhanced framework called Observe, Align, and Enhancement (OAE) for the creation of the radiology report to get around these problems: Observe, which leverages retrieval techniques to enhance visual feature comprehension by identifying similar images and associated reports; Align, where the retrieved contextual reports guide the generation of an initial diagnostic report ensuring semantic consistency; and Enhance, an iterative refinement process that incorporates additional textual information to improve semantic coherence and diagnostic precision. The proposed framework is shown in a comprehensive assessment of the IU-XRAY and MIMIC-CXR datasets, which shows that it performs better than the current state-of-the-art approaches, including improved diagnostic accuracy and reported quality.
- Open Access
GeRe: Towards Efficient Anti-Forgetting in Continual Learning of LLM via General Samples Replay
The continual learning capability of large language models (LLMs) is crucial for advancing artificial general intelligence. However, continual fine-tuning LLMs across various domains often suffers from catastrophic forgetting, characterized by: 1) significant forgetting of their general capabilities, and 2) sharp performance declines in previously learned tasks. To simultaneously address both issues in a simple yet stable manner, we propose General Sample Replay (GeRe), a framework that use usual pretraining texts for efficient anti-forgetting. Beyond revisiting the most prevalent replay-based practices under GeRe, we further leverage neural states to introduce a enhanced activation states constrained optimization method using threshold-based margin (TM) loss, which maintains activation state consistency during replay learning. We are the first to validate that a small, fixed set of pre-collected general replay samples is sufficient to resolve both concerns--retaining general capabilities while promoting overall performance across sequential tasks. Indeed, the former can inherently facilitate the latter. Through controlled experiments, we systematically compare TM with different replay strategies under the GeRe framework, including vanilla label fitting, logit imitation via KL divergence and feature imitation via L1/L2 losses. Results demonstrate that TM consistently improves performance and exhibits better robustness. Our work paves the way for efficient replay of LLMs for the future. Our code and data are available at https://github.com/Qznan/GeRe.
Open AccessGeRe: Towards Efficient Anti-Forgetting in Continual Learning of LLM via General Samples Replay
The continual learning capability of large language models (LLMs) is crucial for advancing artificial general intelligence. However, continual fine-tuning LLMs across various domains often suffers from catastrophic forgetting, characterized by: 1) significant forgetting of their general capabilities, and 2) sharp performance declines in previously learned tasks. To simultaneously address both issues in a simple yet stable manner, we propose General Sample Replay (GeRe), a framework that use usual pretraining texts for efficient anti-forgetting. Beyond revisiting the most prevalent replay-based practices under GeRe, we further leverage neural states to introduce a enhanced activation states constrained optimization method using threshold-based margin (TM) loss, which maintains activation state consistency during replay learning. We are the first to validate that a small, fixed set of pre-collected general replay samples is sufficient to resolve both concerns--retaining general capabilities while promoting overall performance across sequential tasks. Indeed, the former can inherently facilitate the latter. Through controlled experiments, we systematically compare TM with different replay strategies under the GeRe framework, including vanilla label fitting, logit imitation via KL divergence and feature imitation via L1/L2 losses. Results demonstrate that TM consistently improves performance and exhibits better robustness. Our work paves the way for efficient replay of LLMs for the future. Our code and data are available at https://github.com/Qznan/GeRe.
- 2 July 2025
- Published by Springer Nature
- p. 470-484
PVEIN: A Pretrained Vertex Embedding Infer Network for Open-Domain Question Answer Scoring
- 2 July 2025
- Published by Springer Nature
- p. 470-484
PVEIN: A Pretrained Vertex Embedding Infer Network for Open-Domain Question Answer Scoring
- Open Access
- 30 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1293-1302
Local and Global Aware Document Image Enhancement with Residual Denoising Diffusion Model
- 30 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1293-1302
Open AccessLocal and Global Aware Document Image Enhancement with Residual Denoising Diffusion Model
- Open Access
Document Image Rectification Bases on Self-Adaptive Multitask Fusion
Deformed document image rectification is essential for real-world document understanding tasks, such as layout analysis and text recognition. However, current multi-task methods -- such as background removal, 3D coordinate prediction, and text line segmentation -- often overlook the complementary features between tasks and their interactions. To address this gap, we propose a self-adaptive learnable multi-task fusion rectification network named SalmRec. This network incorporates an inter-task feature aggregation module that adaptively improves the perception of geometric distortions, enhances feature complementarity, and reduces negative interference. We also introduce a gating mechanism to balance features both within global tasks and between local tasks effectively. Experimental results on two English benchmarks (DIR300 and DocUNet) and one Chinese benchmark (DocReal) demonstrate that our method significantly improves rectification performance. Ablation studies further highlight the positive impact of different tasks on dewarping and the effectiveness of our proposed module.
Open AccessDocument Image Rectification Bases on Self-Adaptive Multitask Fusion
Deformed document image rectification is essential for real-world document understanding tasks, such as layout analysis and text recognition. However, current multi-task methods -- such as background removal, 3D coordinate prediction, and text line segmentation -- often overlook the complementary features between tasks and their interactions. To address this gap, we propose a self-adaptive learnable multi-task fusion rectification network named SalmRec. This network incorporates an inter-task feature aggregation module that adaptively improves the perception of geometric distortions, enhances feature complementarity, and reduces negative interference. We also introduce a gating mechanism to balance features both within global tasks and between local tasks effectively. Experimental results on two English benchmarks (DIR300 and DocUNet) and one Chinese benchmark (DocReal) demonstrate that our method significantly improves rectification performance. Ablation studies further highlight the positive impact of different tasks on dewarping and the effectiveness of our proposed module.