DeepSeek paper on their new Manifold-Constrained Hyper-Connections (mHC) method could fundamentally change how AI models are trained and scaled—with implications for the entire industry.

☆ Yσɠƚԋσʂ ☆@lemmygrad.ml · 5 months ago

DeepSeek paper on their new Manifold-Constrained Hyper-Connections (mHC) method could fundamentally change how AI models are trained and scaled—with implications for the entire industry.

joaomarrom [he/him, comrade/them]@hexbear.net · 5 months ago

it is January 2025, researchers from Deepseek disrupt the LLM industry

it is January 2026, researchers from Deepseek disrupt the LLM industry

LadyCajAsca [she/her, comrade/them]@hexbear.net · 5 months ago

China keeps winning

BB84@mander.xyz · 5 months ago

this is not their website. it’s an SEO squatter trying to make ad money. the article is crappy slop explanation. the real research article is at https://arxiv.org/abs/2512.24880

@yogthos@lemmygrad.ml, please consider changing the link in your post

DeepSeek paper on their new Manifold-Constrained Hyper-Connections (mHC) method could fundamentally change how AI models are trained and scaled—with implications for the entire industry.

DeepSeek paper on their new Manifold-Constrained Hyper-Connections (mHC) method could fundamentally change how AI models are trained and scaled—with implications for the entire industry.

DeepSeek AI 2026: R1, V3 & Local AI Hub