“In our quest to push the limits of SenseNova’s capabilities, SenseTime remains guided by the law of scaling as we rely on our grand (language) model based on this three-level architecture: knowledge, reasoning and execution,” Li said, referring to expanding the size of the LLM by increasing its number of parameters to increase its learning and generalization ability.
Given the same insight, SenseNova 5.0 provides better insight, summary, and Q&A to support vertical applications such as in the education and content industries, according to a statement from SenseTime. The updated LLM is also touted as having “best-in-class mathematical, coding and reasoning capabilities, providing a strong foundation for applications in finance and data analytics”.
SenseNova 5.0 also supports analyzing and understanding high-definition images, as well as generating text in images, according to SenseTime. Additionally, the latest iteration of the LLM also extracts complex data across documents and summarizes answers to questions, “providing strong multimodal interaction capability.”
The updated LLM, according to SenseTime, underwent more than 10 terabytes of tokenization, covering a large amount of synthetic data, which improved its pop-up coverage – referring to the strip of text that a model d AI can process during a conversation with users. – at around 200,000.
