CSIG Enterprise Visit to Qihoo 360: Multimodal and Cross‑Modal Learning in the Era of Large Models

On June 29, 2023, the China Society of Image and Graphics (CSIG) organized the "CSIG Enterprise Visit – Into Qihoo 360" event, co‑hosted by Beijing Qihoo Technology Co., Ltd. and the CSIG Youth Working Committee, attracting more than 1,000 participants both online and offline.

The theme of the event was "Multimodal and Cross‑Modal Learning in the Era of Large Models," with invited experts and scholars from Peking University, Zhejiang University, Harbin Institute of Technology, and Wuhan University discussing cutting‑edge trends alongside the Qihoo 360 technical team.

Figure 1: Event venue

CSIG Vice‑Secretary General and Beijing University of Aeronautics and Astronautics professor Liu Si delivered an opening speech, introducing the society’s history, structure, academic activities, industry‑academic‑research integration, and talent recommendation programs, and thanked Qihoo 360 for its strong support.

Qihoo 360 Vice President and Head of the Technical Platform, Yin Yuhui, welcomed the guests, emphasizing that multimodal large models are the hottest research focus in both academia and industry and a necessary step toward general artificial intelligence. He outlined Qihoo 360’s dual‑track AI development strategy—strengthening core capabilities while pursuing application scenarios—and introduced the latest progress of the "Intelligent Brain 4.0" project.

During the academic report session, scholars from universities and research institutes presented their recent work:

Zhejiang University Professor Zhao Zhou presented "Cross‑Modal Audio‑Video Generation Model Research," covering NATSpeech for real‑time high‑quality lightweight speech synthesis, DiffSinger for high‑performance multi‑task singing synthesis, AudioGPT for open, temporal, multi‑task audio generation, and GeneFace for controllable, robust, multimodal facial video synthesis.

Wuhan University Professor Wu Yu delivered a talk on "Multimodal Perception and Generation," sharing advances in multimodal learning such as audio‑video understanding, vision‑language perception models, and diffusion‑based multimodal generation methods.

Harbin Institute of Technology (Shenzhen) Associate Professor Zhang Zheng discussed "Multi‑Source Interactive Emotion Understanding and Analysis," focusing on facial, speech, and multimodal emotion feature extraction, cross‑domain emotion analysis, and multimodal emotion recognition algorithms.

Peking University Assistant Researcher He Xiangteng presented "Fine‑Grained Cross‑Media Classification and Retrieval," reviewing the current state of research and future directions in this field.

Dr. Leng Dawei, Head of Vision Engine at Qihoo 360 AI Research Institute, gave the keynote "Multimodal and Cross‑Modal Learning in the Era of Large Models," reviewing recent progress in MLLM (Multimodal Large Language Model) research, analyzing native multimodal versus single‑modal expert stitching approaches, and outlining Qihoo 360’s R&D thoughts, recent achievements, and future plans.

Figure 2: Professor Zhao Zhou presenting

Participants also toured Qihoo 360’s exhibition hall, gaining insight into the company’s development history, strategic positioning, contributions to national cyber‑security, and ongoing efforts in digital security, digital transformation, and artificial intelligence.

Figure 3: Tour of Qihoo 360 facilities

The event concluded successfully, fostering interaction between academia, research institutes, and industry. CSIG plans to continue promoting industry‑academic‑research integration, building platforms for technology and talent exchange, and strengthening collaborations to advance the image and graphics field.

Organizer Introduction

Founded in 2015, the Qihoo 360 AI Research Institute focuses on cutting‑edge computer vision, deep natural language understanding, speech‑semantic interaction, large‑scale deep learning, and robotic motion. Its technologies are applied to smart IoT, intelligent security big data, internet information distribution, enterprise digitalization, and intelligent vehicles. The team has achieved top results in international competitions, contributed to national and municipal key projects, helped build a national big‑data engineering laboratory, and developed a security brain selected for the national next‑generation AI open‑innovation platform. Members hold degrees from leading universities worldwide and have experience at major tech companies.