Lingyu Zhou

Lingyu Zhou

PhD Student

Sichuan University

Research Interests

Hallucination Mitigation in Multimodal LLMs
Visual Evidence Grounding
Radiology Report Generation
Medical Image Analysis

About

I am a PhD student at the College of Computer Science, Sichuan University, advised by Prof. Zhang Yi (Foreign Member of the Russian Academy of Engineering, IEEE Fellow). I am a member of the Machine Intelligence Lab.

Prior to this, I obtained my B.Eng. degree in Computer Science and Technology from Sichuan University in 2022.

My research centers on mitigating hallucinations in Multimodal Large Language Models (MLLMs) within real-world clinical scenarios. Specifically, I investigate how to effectively embed visual evidence grounding with rich semantic context into LLMs, enabling faithful and interpretable cross-modal reasoning. My work spans radiology report generation, medical image segmentation, and computer-aided diagnosis, aiming to build trustworthy AI systems that bridge the gap between multimodal intelligence and clinical practice.

Selected Publications

View All

Mitigating Entity Hallucinations in 3D Radiology Report Generation via Dual-Stream Alignment

Lingyu Zhou, Yue Yu, Zhang Yi, Xiuyuan Xu

Proceedings of the AAAI Conference on Artificial Intelligence (AAAI)

Proposed DEAR framework to mitigate entity hallucinations in 3D radiology report generation through dual-stream alignment. AAAI 2026 Poster, CCF-A.

Glance Before You Tell: Anomaly-Guided 3D Radiology Report Generation with Heat-Conduction Slice Encoders

Lingyu Zhou, Dingwen Pi, Zhang Yi, Jun Wu, Xiuyuan Xu

Proceedings of the Conference on Neural Information Processing Systems (NeurIPS)

Proposed HeatRAD, a glance-before-tell framework that routes anomaly-guided slice evidence into an organ-ordered visual prefix for 3D radiology report generation. NeurIPS 2026, CCF-A, Under Review.

Navigate Before You Generate: Volume Evidence Grounding for 3D Radiology Reports

Lingyu Zhou, Dingwen Pi, Jun Wu, Zhang Yi, Xiuyuan Xu

Proceedings of the Conference on Neural Information Processing Systems (NeurIPS)

Proposed Vol-Nav, a volume evidence grounding paradigm for 3D radiology report generation. NeurIPS 2026, CCF-A, Under Review.

ForeMatch: Rethinking Foreground Consistency for Efficient Semi-Supervised 3D Medical Image Segmentation

Lingyu Zhou, Zhengyang Xu, Zhang Yi, Deng Xiong, Xiuyuan Xu

Proceedings of the ACM International Conference on Multimedia (ACM MM)

Proposed ForeMatch for efficient semi-supervised 3D medical image segmentation via foreground consistency. ACM MM 2026, CCF-A.

Efficient and Gender-Adaptive Graph Vision Mamba for Pediatric Bone Age Assessment

Lingyu Zhou, Zhang Yi, Kai Zhou, Xiuyuan Xu

Proceedings of the International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)

Proposed GGVMamba, a gender-adaptive Graph Vision Mamba framework for pediatric bone age assessment. MICCAI 2024 Poster, CCF-B.

News

2026-07
One paper accepted at ACM MM 2026! 🎉
2026-06
Led team to win First Prize at Sichuan Province "Challenge Cup" Competition! 🏆
2026-05
Two papers submitted to NeurIPS 2026.
2026-05
One paper narrowly missed acceptance at ICML 2026.
2026-03
One paper submitted to ACM MM 2026.
2026-01
One paper submitted to ICML 2026.
2025-11
One paper accepted as Poster at AAAI 2026! 🎉