An Open Source Mathematical AI Model, Dipic Prover V2 released

The Chinese -based artificial intelligence (AI) firm, The Hangzhou, Deepesic, released a latest version of his Power model on Wednesday. Dubbed Deep Sak-Purver-V2, it is a very special model that focuses on proving formal mathematical ideas. The large language model (LLM) uses lean 4 programming language to check whether mathematical evidence is logically compatible by analyzing each step independently. Like the previous release of a Chinese firm, Dipic-Purver-V2 is an open source model and can be downloaded from popular reservoirs such as gut hub and throat faces.

DiPsic’s new math -based AI model is here

AI firm told the new model on his gut hub in detail Listing Page It is primarily a model of theoretical chains that has a visible chain off -thought (COT), which works in the domain of mathematics. It is made from the Dipici-V3AI model and has been erected, which was released in December 2024.

DEP Sec-Prover-V2 can be used in different ways. It can solve high school with college -level mathematics issues and find and solve mistakes in the evidence of mathematics theory. It can also be used as teaching aid and can produce step -step -by -step explanations of evidence, and it can help mathematicians and researchers to find new ideas and prove their authenticity.

It is available in two models sizes – seven billion parameters size and more than 671 billion parameter size. Although the latter is trained at the top of the Dipic-V 3 Base, the former is made at the Dipseic-Prover-V-1.5 Base and it comes up to 32,000 tokens in length.

Before training, researchers implemented a cold start training system, indicating the twenty models, indicating complex issues. These issues worked as a series of subgolis. Subsequently, evidences of resolved subgals were added to the COT and found the base model to create an early cold start to learn reinforcement.

Specifically, in addition to the gut hub, the AI ​​model can also be downloaded from the Facial Face of Depsek. Listing. The Prover-V2 model highlights how changes in the training process of AI models can significantly improve their special ability. Like the other open source model release, details about basic architecture or major dataset are not known.