Hello 👋

I am a Ph.D. candidate at the Institute of Computing Technology, Chinese Academy of Sciences (ICT, CAS), advised by Prof. Liang Li. Before that, I completed my Master’s program at Shandong University, working with Prof. Yuankai Qi.

My research focuses on multimodal understanding and generation, with an emphasis on audiovisual world models and multimodal speech generation. I explore AI dubbing for animated and live-action characters, fine-grained control through natural-language instructions, and cross-lingual voice identity preservation, aiming to enable trustworthy AI-driven film production. My research directions include:

• Reinforcement Learning: aligning speech generation models with human preferences through multi-objective reinforcement learning and perceptual feedback, balancing linguistic accuracy, audiovisual coherence, and expressive quality.
• Audiovisually Synchronized V2C Models: exploring autoregressive and non-autoregressive approaches to model the relationships among language, lip movements, facial expressions, and speech prosody, enabling speech generation with precise audiovisual alignment.
• High-Fidelity AI Dubbing: developing two-stage pretraining frameworks that transfer knowledge from text-to-speech (TTS) pretraining to instruction-guided dubbing, with a focus on improving speech quality and naturalness.
• Unlearning for Generative Models: investigating how to selectively remove the influence of designated training data and suppress unwanted learned associations while preserving the models’ broader generative capabilities, supporting responsible model development and deployment.

I am looking for collaborators in audio, speech, and multimodal AI—from generation and understanding to practical systems. Feel free to reach out to me if you have an interest in the relevant topics.