# Summary of the Article:
– Large Language Models (LLMs) have shown progress in natural language processing tasks.
– Challenges still exist in achieving robust reasoning with LLMs, including the need for supervised fine-tuning.
– Researchers are working on new approaches to address issues like poor readability and computational efficiency while improving reasoning capabilities.
## DeepSeek AI’s Contribution:
– DeepSeek AI has released DeepSeek-R1-Zero and DeepSeek-R1, reasoning models that promote reasoning capability in LLMs through reinforcement learning.
### Author’s Take:
DeepSeek AI’s launch of DeepSeek-R1-Zero and DeepSeek-R1, the first-generation reasoning models, signals a shift towards incentivizing reasoning capability in Large Language Models (LLMs) through reinforcement learning. This advancement comes as a response to challenges in scalability, generalization, and computational efficiency faced by LLMs in achieving robust reasoning. The research community’s exploration of new approaches showcases a commitment to overcoming existing obstacles and pushing the boundaries of natural language processing technology.
Click here for the original article.