Renowned AI Scholars from SJTU, CUHK (Shenzhen) and Tencent Hunyuan to Present at MLNLP 2026 Symposium

Event Overview The MLNLP community, a prominent international machine learning and natural language processing forum, is hosting its 41st academic symposium online on 2026‑07‑26 from 09:00 to 12:00 Beijing time. The event is co‑organized by the MLNLP community, the Chinese Information Processing Society of China Youth Working Committee, and the Large Model & Generation Professional Committee, with strategic partnership from Tencent.

Schedule The symposium is divided into two halves. The first half is chaired by Assistant Professor Zhang Linfeng (Shanghai Jiao Tong University), and the second half by Assistant Professor Pan Ling (Hong Kong University of Science and Technology). Detailed agenda information is provided on the event page.

Speakers and Talks

Zhang Linfeng – Assistant Professor, School of Artificial Intelligence, Shanghai Jiao Tong University. Talk: “Lifelong Learning: Forgetting Resistance, Safety, and Personalization.” Abstract: Discusses challenges of continual learning for AI agents, proposes context‑aware parameter decomposition to mitigate knowledge forgetting, safety‑aware constrained optimization to preserve alignment, and memory‑based methods for long‑term multimodal interaction and user‑specific personalization.

Pan Ling – Assistant Professor, Department of Electronic and Computer Engineering, Hong Kong University of Science and Technology. Talk: “Welcome to the Age of Self‑Evolving Agents.” Abstract: Introduces self‑evolving agents, analyzes limitations of large models in domain‑specific scenarios, presents techniques for autonomous model evolution, and demonstrates performance gains in Chinese GPU operator generation.

Yang Yibo – Associate Professor, School of Artificial Intelligence, Shanghai Jiao Tong University. Talk: “Lifelong Learning: Forgetting Resistance, Safety, and Personalization.” (same title as Zhang’s talk, detailed abstract above).

Liu Zhen – Assistant Professor, School of Data Science, The Chinese University of Hong Kong (Shenzhen). Talk: “Efficient Fine‑Tuning of Vision Generative Models.” Abstract: Explores reward‑gradient based fine‑tuning, preference‑guided optimization for diffusion models, and drift‑model methods for single‑step generators, aiming to improve post‑training efficiency and quality.

Chen Jianghai – Senior Researcher, Tencent Hunyuan. Talk: “HunYuan UniRL: A Unified Multimodal Reinforcement Learning Framework.” Abstract: Presents UniRL, an open‑source framework that integrates RL post‑training for large language, vision‑language, diffusion, and unified multimodal models, supporting distributed training back‑ends, multi‑modal reward signals, and novel Flow‑DPPO/DRPO algorithms, with full compatibility to models such as SD‑3/3.5, FLUX‑2‑Klein, HunyuanVideo, Qwen‑3.

Registration Participants can register by scanning the QR code provided in the announcement; registration details and the live‑stream links (Bilibili and WeChat Video) are shared in the MLNLP community group.