DeepSeek Enhances its Math-Oriented AI Model, Prover
The Chinese AI lab DeepSeek has quietly enhanced Prover, its AI model designed to solve mathematical proofs and theorems.
As reported by the South China Morning Post, DeepSeek uploaded the latest version of Prover, V2, along with a distilled variant, to the AI development platform Hugging Face late on Wednesday. The new version seems to be based on the startup’s V3 model, which features 671 billion parameters and utilizes a mixture-of-experts (MoE) architecture.
“Parameters typically drive a model’s capability to solve problems, while MoE (Mixture of Experts) breaks down tasks into smaller subtasks and actively assigns them to specialized ‘expert’ components.”
DeepSeek last updated Prover in August, characterizing it then as a custom, publicly accessible AI model designed for formal theorem proving and mathematical reasoning.
In February, Reuters reported that DeepSeek was reportedly exploring the possibility of raising external funding for the first time. “The company has recently launched an upgraded version of V3, a general-purpose model, and is set to update its R1 ‘reasoning’ model soon.”
Read the original article on: Techcrunch
Leave a Reply