Sung Kim

sungkim.bsky.social

17 February 2025 16:54

Alibaba (Ant Group)'s LLaDA-8B, a diffusion model with an unprecedented 8B scale LLaDA is a masked diffusion model that follows standard pretraining and SFT while sampling via diffusion. During pretraining, it masks all tokens randomly at ratio; in SFT, only response tokens may be masked.

Owen Winter

owenwntr.bsky.social

16 February 2025 08:55

How are Germans planning to vote? I took a look, with five charts and a bunch of data 🇩🇪📈

Owen Winter

owenwntr.bsky.social

16 February 2025 08:55

Among others, two key factors stand out. Firstly, age—the number one predictor of voting intentions. While older voters stick with the traditional big parties, young and working age Germans are split between ideologically diverse smaller parties

mr. TIM

timkellogg.me

13 February 2025 19:18

ngl i’m really excited about this

Lionel Page

lionelpage.bsky.social

13 February 2025 19:18

The paradox of power: The position of a hegemonic power is, paradoxically, often based not on raw power alone, but on the ability to convince weaker players that this power will not be used against them. www.optimallyirrational.com/p/the-parado...