PhD student of GSAI attended the 18th European Conference on Computer Vision (ECCV 2024)
In October 2024, our PhD student Yifan Li traveled to Milan, Italy, to attend the 18th European Conference on Computer Vision (ECCV 2024). During the conference, he delivered a group presentation and participated in a poster session, showcasing her research findings to attending experts. As a top-tier international conference in the field of computer vision, ECCV covers cutting-edge research topics such as multimodal large language models, generative models, and autonomous driving. It serves as a vital platform for academics, industry professionals, and business experts worldwide to share the latest technological advancements, present innovative applications, and discuss future development trends.
Yifan Li: This trip to Milan was an incredibly memorable experience. At the conference, I presented a paper on the safety alignment of multimodal large language models, which evaluate potential safety vulnerabilities in existing models. Based on our findings, we proposed a jailbreak method called HADES, which effectively exploits weaknesses in both open-source and closed-source multimodal large language models. We also conducted preliminary explorations on enhancing model defenses. On the last day of the conference, I delivered an oral presentation and a poster session, discussing my work with researchers in the same field. It was deeply fulfilling to introduce my work and gain recognition from others.
During the rest of the conference, I attended oral presentations from other fields and browsed posters of papers that piqued my interest. Engaging in face-to-face discussions with authors was far more dynamic than simply reading papers on a screen. Through these exchanges, I connected with peers working in related research areas, refined my future research plans, and gained new insights into fields I had previously explored less.