Kunlun Wanwei Technology Co., Ltd. has announced the official launch of the invitation testing for its "Skywork o1" model (Skywork o1). This news marks an important step for the company in the field of general artificial intelligence, aiming to achieve higher levels of AI applications through technological innovation.
Skywork o1 is a series of models released by Kunlun Wanwei Group, featuring slow thinking and reasoning capabilities. It is the first o1 model in China with logical reasoning abilities in Chinese. This model not only embodies thinking, planning, and reflection in its outputs but also shows significant improvements in reasoning capabilities compared to ordinary models on standard evaluation sets. The released Skywork o1 includes three models: Skywork o1Open, Skywork o1Lite, and Skywork o1Preview, where Skywork o1Open is an open-source model based on Llama3.18B, Skywork o1Lite has complete thinking abilities, and Skywork o1Preview offers a more diverse and in-depth thinking process.
Kunlun Wanwei Group stated that Skywork o1Open has significantly improved in mathematical and coding metrics, elevating the performance of Llama-3.1-8B to the SOTA level within the same ecosystem, and unlocking complex mathematical tasks that many lightweight models cannot solve. In addition, the company will open source two Process Reward Models (PRM) for reasoning tasks: Skywork o1Open-PRM-1.5B and Skywork o1Open-PRM-7B, which can score each step and enhance reasoning and thinking abilities.
The performance improvement of Skywork o1 in logical reasoning tasks is attributed to Kunlun Wanwei's self-developed training programs, including reasoning reflection training, reinforcement learning for reasoning abilities, and reasoning planning. The application of these technologies enables Skywork o1 to excel in problems related to common sense reasoning, logical reasoning, mathematical reasoning, and ethical decision-making.
Kunlun Wanwei emphasizes that the launch of Skywork o1 is not only an important initiative in the company's "All in AGI and AIGC" strategy but also a significant step in building an AI technology stack. The company will continue to strive for the realization of general artificial intelligence, allowing everyone to better shape and express themselves.
Taste it here:www.tiangong.cn
