yogthos

joined 5 years ago
[–] yogthos@lemmy.ml 14 points 3 months ago

Because people doing work need to pay bills and make a living? You also seem to be confused about what capitalism is. A worker selling their own labour is not capitalism. Capitalism is when you use capital to buy up the means of production and then hire workers to work for you and appropriate the value they produce.

[–] yogthos@lemmy.ml 15 points 3 months ago (4 children)

Imagine living in year 2025 and not being able to figure out how to find a PDF online. Explains why you're such an ignoramus I guess.

[–] yogthos@lemmy.ml 14 points 3 months ago

you gotta love comments from the dummy.world instance

[–] yogthos@lemmy.ml 15 points 3 months ago (8 children)

Great, plenty of other economists have concluded the opposite. Here are a few books you can read to educate yourself.

[–] yogthos@lemmy.ml 8 points 3 months ago (42 children)

lmfao China does not embrace capitalism even in the slightest, meanwhile Russia's public sector is almost as big as China's

[–] yogthos@lemmy.ml 10 points 3 months ago

yup... eblows down, bottoms up

[–] yogthos@lemmy.ml 1 points 3 months ago

I'm incredibly excited about Asahi Linux, and it's amazing to see just how much progress they've made with it. M series architecture is strictly superior to anything else that's currently available, and being able to run Linux on it is really great. The only thing that I miss with it is hibernation, but that's really not that big of a deal when it comes to day to day usage.

[–] yogthos@lemmy.ml 6 points 3 months ago
[–] yogthos@lemmy.ml 14 points 3 months ago

Let's just look at actual academic studies on USSR

Professor of Economic History, Robert C. Allen, concludes in his study without the 1917 revolution is directly responsible for rapid growth that made the achievements listed above possible:

Study demonstrating the steady increase in quality of life during the Soviet period (including under Stalin). Includes the fact that Soviet life expectancy grew faster than any other nation recorded at the time:

A large study using world bank data analyzing the quality of life in Capitalist vs Socialist countries and finds overwhelmingly at similar levels of development with socialism bringing better quality of life:

This study compared capitalist and socialist countries in measures of the physical quality of life (PQL), taking into account the level of economic development.

This study shows that unprecedented mortality crisis struck Eastern Europe during the 1990s, causing around 7 million excess deaths. The first quantitative analysis of the association between deindustrialization and mortality in Eastern Europe.

Romania, the inustrialization of an agrarian economy under socialist planning

Then, we can look at how do people who lived under communism feel now that they got a taste of capitalism?

[–] yogthos@lemmy.ml 6 points 3 months ago

Uh no, other systems very much do exist. Go look at China or Vietnam as two obvious counter examples. The reality is that the US will continue on the exact same trajectory regardless of which oligarch owned party is in charge. The US is a dictatorship of capital. Until people start organizing outside voting and create an actual labour movement, nothing will fundamentally change.

 
 
 

This paper introduces DiffuCoder, a 7B-scale open-source masked diffusion large language model (dLLM) specifically designed for code generation.

The research provides insights into how dLLMs generate content, distinguishing their decoding behavior from that of autoregressive (AR) models. Unlike AR models, dLLMs can intrinsically adjust their generation causality and increasing sampling temperature diversifies not just token choices but also their generation order, creating a rich search space for reinforcement learning (RL).

This flexibility allows dLLMs to be more non-autoregressive and generate tokens in a less sequential, more "human-like" code writing manner.

To leverage this diversity and improve performance, the paper proposes coupled-GRPO RL algorithm. This method utilizes a coupled-sampling scheme that constructs complementary mask noise during training to reduce the variance of token log-likelihood estimates while maintaining training efficiency.

Experimentally, coupled-GRPO significantly boosts DiffuCoder's performance on code generation benchmarks, notably improving EvalPlus scores by 4.4% with training on only 21K samples. The research also shows that coupled-GRPO trained models experience a smaller performance drop when decoding steps are halved (resulting in a 2x speedup), indicating increased parallelism and reduced reliance on AR bias during decoding.

available at https://huggingface.co/apple/DiffuCoder-7B-cpGRPO

view more: ‹ prev next ›