Zhongyu Yang

About Me

I am a Research Intern at Tencent Hunyuan (青云人才计划 / Qingyun Top Talent Program), where my research centers on omni-video understanding — jointly interpreting a video together with its accompanying audio and language (e.g., Script-a-Video, deep structured audio-visual captioning). Previously, I was a remote research intern at Vision-CAIR, KAUST, advised by Mohamed Elhoseiny, and a research intern at SenseTime. I received my B.S. in Mathematics (minor in Management) from Lanzhou University.

My research seeks to advance multimodal models from surface-level recognition toward genuine understanding — reasoning about why events occur and what follows, with inferences grounded in evidence that is consistent across modalities and over time. My work spans four interconnected directions:

Research Interests

• Omni-Video Understanding: long-form, audio-visual, and tool-augmented video comprehension that integrates vision, audio, and language over time (SVAgent_CVPR, ParaVT_Tech Report, Script‑a‑Video_Tech Report)
• Compositional Multimodal Reasoning: evaluating and strengthening how models exploit cross-modal evidence when reasoning over evolving world states (UFO_ICML, InEx_AAAI, MERMAID_EMNLP, XR_WWW)
• Generative Modeling as Evidence: leveraging generated images, sequences, and future states as intermediate evidence for understanding and prediction (WikiAutoGen_ICCV, ReChar_{SIGGRAPH Asia}, Evolving Visual Generation_Tech Report, CHAT_ECCV)
• Efficient & Trustworthy Models: token compression and safety-oriented mechanisms that keep large multimodal and language models scalable and reliable (Script_TMLR, SCOPE_KDD)

In the long term, I aim to develop general-purpose multimodal systems that perceive, reason, and communicate across vision, audio, language, and action in dynamic, real-world environments.

I am always open to research collaborations and discussions — please feel free to reach out.

News

• [Pinned to top] I am actively seeking a PhD position and a research internship in 2026. I am always happy to discuss potential opportunities!
• [2026.06] CHAT and MultiHaystack were accepted to ECCV 2026 🎉.
• [2026.05] SCOPE was accepted to KDD 2026 🎉.
• [2026.05] UFO was accepted to ICML 2026 🎉. The dataset and code are now open-sourced.
• [2026.02] EmoRes was accepted to ACL 2026 🎉.
• [2026.02] SVAgent and MeteoNet were accepted to CVPR 2026 🎉.
• [2026.01] XR was accepted to WWW 2026 🎉.
• [2025.11] InEx was accepted to AAAI 2026 🎉.
• [2025.10] Script was accepted to TMLR 2025 🎉.
• [2025.08] ReChar was accepted to SIGGRAPH Asia 2025 🎉.
• [2025.08] MERMAID was accepted to EMNLP 2025 (Main) 🎉.
• [2025.06] WikiAutoGen was accepted to ICCV 2025 🎉.
• [2025.03] We present WikiAutoGen for automated multimodal Wikipedia-style article generation. Featured in Hugging Face Daily Papers and reposted by AK.
• [2024.12] Started Remote Research Internship with Jun Chen and Mohamed Elhoseiny (KAUST), focusing on web-scale RAG systems for vision.
• [2024.11] Released ReChar: structure-preserving and user-aesthetic-enhanced character generation.
• [2024.05] Our paper Green Effect of Energy Transition Policy was accepted to Finance Research Letters (SSCI Q1 Top journal) 🎉.
• [2024.04] National Innovation Project funded: FPGA-Based AI Doctor, advised by Prof. Xinhua Wang.
• [2024.04] Joined CUHKSZ as Research Assistant with Ruimao Zhang, focusing on AI4Science and CV.
• [2024.03] Began Remote Research Internship with Yingfang Yuan at Heriot-Watt University on AIGC.
• [2024.01] Project UNet-Centric MambaMorph selected as Outstanding Undergraduate Project (TOP 0.1%) under Barley Plan, advised by Prof. Wenting Zhang.
• [2023.05] Paper Environmental Quality in OECD Countries was accepted to Renewable Energy (SCI Q1 Top journal) 🎉.
• [2023.03] As sophomore, led innovation project on Tropical Linear Representation of Chinese Monoids, advised by Prof. Wenting Zhang. [Tech Report] can be finding.

Experience

• Research Intern, Tencent Qingyun Top Talent Program, Hunyuan, Tencent Mar. 2026 – Present
• Research Intern, General Perceptual Computing Group, SenseTime Feb. 2025 – Feb. 2026
• Remote Research Intern, BCML Lab, Heriot-Watt University Mar. 2024 – Present
• Remote Research Intern at Vision-CAIR, KAUST Dec. 2024 – Jan 2026
• Research Assistant at LIAS Lab, CUHK (Shenzhen) Apr. 2024 – Nov. 2024
• Data Analysis Assistant, iFLYTEK Jun. 2023 – Aug. 2023

Selected Publications [Google Scholar]

Services

Conference Reviewers

• CVPR, ECCV, ICCV, ICLR, NeurIPS, ICML, ACL, EMNLP, AAAI, SIGGRAPH Asia, WWW, KDD

Journal Reviewers

• IJCV, CVIU, TMM, TIP

Powered by Jekyll and Minimal Light theme.