AI for Science
Practical walk through from the world leading research insistution: Shanghai AI Lab
Huanjun Kong
Session time: 2025.07.24 2pm-4pm
Shanghai Artificial Intelligence Laboratory — https://www.shlab.org.cn/
Huanjun Kong graduated from the University of Science and Technology of China and spent 10 years in industry, working across computer vision and natural language processing. Since Oct 2024 he has focused on LLM for Science, leading a team that has produced several world‑class contributions in Shanghai AI Lab.
Key Outputs in the past year
SeedLLM‑Rice: Molecular Plant (Cell Press) — https://www.cell.com/molecular-plant/fulltext/S1674-2052(25)00172-8. Rice‑specific LLM integrated with a biological knowledge graph.
SeedBench: ACL 2025 Main — https://github.com/open-sciencelab/SeedBench. Multi‑task benchmark for evaluating LLMs in seed science.
GraphGen: Pre‑print, under review for EMNLP — https://github.com/open-sciencelab/GraphGen Knowledge‑driven synthetic data generation for superior fine‑tuning.
ROGRAG: ACL 2025 Demo — https://github.com/tpoisonooo/ROGRAG. Robustly optimised Graph‑based Retrieval‑Augmented Generation framework.
Beyond these flagship outputs, Huanjun has open‑sourced the entire supporting codebase and actively maintains a cohesive, end‑to‑end toolchain—covering document processing (MinerU from Shanghai AI Lab), synthetic data generation (GraphGen), benchmarking (SeedBench), domain‑specific LLM training (SeedLLM‑Rice), and robust GraphRAG deployment—enabling researchers in any data‑rich scientific field to adopt and extend it for their own AI‑for‑Science breakthroughs.
Huanjun will open the session with an overview of this journey and share practical stories about how these results were achieved, aiming to inspire comparable advances in other domains. The talk will conclude with an open Q&A—bring your questions and dive into the details.
Last updated