DeepSeek is working with Tsinghua University on reducing the training its AI models need in an effort to lower operational costs.<\/p>\n\n\n\n
The Chinese startup, which roiled markets with its low-cost reasoning model that emerged in January, collaborated with researchers from the Beijing institution on a paper<\/a> detailing a novel approach to reinforcement learning to make models more efficient.<\/p>\n\n\n\n The new method aims to help artificial intelligence models better adhere to human preferences by offering rewards for more accurate and understandable responses, the researchers wrote. <\/p>\n\n\n\n Reinforcement learning has proven effective in speeding up AI tasks in narrow applications and spheres. <\/p>\n\n\n\n However, expanding it to more general applications has proven challenging \u2014 and that\u2019s the problem that DeepSeek\u2019s team is trying to solve with something it calls self-principled critique tuning. <\/p>\n\n\n\n The strategy outperformed existing methods and models on various benchmarks and the result showed better performance with fewer computing resources, according to the paper.<\/p>\n\n\n\n DeepSeek is calling these new models DeepSeek-GRM \u2014 short for \u201cgeneralist reward modeling\u201d \u2014 and will release them on an open source basis, the company said. <\/p>\n\n\n\n Other AI developers, including Chinese tech giant Alibaba Group Holding and San Francisco-based OpenAI, are also pushing into a new frontier of improving reasoning and self-refining capabilities while an AI model is performing tasks in real-time.<\/p>\n\n\n\n