Searching for the ImageNet of the financial sector - Industry and academic dialogue behind the first multimodal credit assessment benchmark.

On February 5th, Qifu Technology. Inc ADR Class A (QFIN.US) hosted a live discussion focused on industry and academic frontiers on the topic of "Establishing Standards for Multi-Modal AI in Credit Assessment". During the live broadcast, the first multi-modal evaluation benchmark for credit scenes, FCMBench-V1.0, recently released by researchers from Qifu Technology. Inc ADR Class A in conjunction with Fudan University and South China University of Technology, became the core topic. This benchmark, derived from real credit business scenarios, is designed to evaluate tasks around key processes such as multi-modal perception, reasoning, and decision-making, and concurrently open-source datasets and evaluation tools, aiming to establish a widely accepted "ruler" for financial AI. During the discussion, three guests from the frontlines of the industry and academic forefront addressed the same issue from different perspectives: without unified standards, it is difficult for financial AI to truly take off. Dr. Yang Yehui, Head of Multi-Modal at Qifu Technology. Inc ADR Class A, began with a metaphorical comparison between AI and application scenarios, emphasizing that AI is a tool while industries like finance and healthcare are fertile ground. He stressed the importance of having a unified and fair evaluation system to prevent decision-making from losing focus. The value of FCMBench lies in leveling the playing field for models and testing their abilities in real business conditions. Dr. Xu Yanwu, a professor at South China University of Technology, offered another viewpoint from cross-industry experiences. He pointed out that AI has long been involved in insurance pricing, asset assessment, and quantitative trading, even though its value may not be immediately visible in consumer-facing products. In conclusion, the host expressed the importance of industry, academia, and research institutions continuing to collaborate for the development of financial AI. She invited more partners to participate in dataset testing, evaluation, and competitions to help shape the "ImageNet of the financial industry" through cooperation and consensus.