On: DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL

[wiether]

Same here.

Following all the hype I tried it on my usual tasks (coding, image prompting...) and all I got was extra-verbose content with lower quality.