During the Hong Kong FinTech Festival, Ant Group Digital Technology took the stage to introduce its groundbreaking 'Multilingual Multimodal Large Model Training Framework' technology. At the heart of this innovation lies an ingenious language-aware optimization framework. This framework boasts the remarkable ability to comprehend and process less widely spoken languages on a profound level.

Test results have been highly promising. The framework has managed to boost accuracy by around 9.5% in mainstream multilingual visual question answering benchmark tests. What's more, on certain tasks, its performance has even outshone that of internationally renowned mainstream closed-source models.

In addition to this, Ant Group Digital Technology has also unveiled an image security framework. This framework is adept at swiftly pinpointing issues within images and delivering interpretable analysis. By doing so, it significantly strengthens risk control measures for digital content.

These two cutting-edge technologies have already been implemented on a large scale in ZOLOZ's document recognition and verification products. They currently support an impressive 119 languages and cater to a wide array of scenarios.