@ariveram2111.bsky.social
ML engineer | Professor | Tech speaker
#AI #ML #MachineLearning #Software #tech
Embracing new challenges | Tech-driven mindset 🚀
#AI #ML #MachineLearning #Software #tech
Embracing new challenges | Tech-driven mindset 🚀
But the most remarkable part from my point of view? 🤯
DeepSeek-R1-Zero is trained with almost pure RL, without supervised fine-tuning (SFT) as a preliminary step. The result? Outstanding reasoning capabilities with minimal labeled data
#AI #ML #MachineLearning #DL #DeepLearning
DeepSeek-R1-Zero is trained with almost pure RL, without supervised fine-tuning (SFT) as a preliminary step. The result? Outstanding reasoning capabilities with minimal labeled data
#AI #ML #MachineLearning #DL #DeepLearning
January 20, 2025 at 5:53 PM
But the most remarkable part from my point of view? 🤯
DeepSeek-R1-Zero is trained with almost pure RL, without supervised fine-tuning (SFT) as a preliminary step. The result? Outstanding reasoning capabilities with minimal labeled data
#AI #ML #MachineLearning #DL #DeepLearning
DeepSeek-R1-Zero is trained with almost pure RL, without supervised fine-tuning (SFT) as a preliminary step. The result? Outstanding reasoning capabilities with minimal labeled data
#AI #ML #MachineLearning #DL #DeepLearning
But the most remarkable part from my point of view? 🤯
DeepSeek-R1-Zero is trained with almost pure RL, without supervised fine-tuning (SFT) as a preliminary step. The result? Outstanding reasoning capabilities with minimal labeled data
DeepSeek-R1-Zero is trained with almost pure RL, without supervised fine-tuning (SFT) as a preliminary step. The result? Outstanding reasoning capabilities with minimal labeled data
January 20, 2025 at 5:35 PM
But the most remarkable part from my point of view? 🤯
DeepSeek-R1-Zero is trained with almost pure RL, without supervised fine-tuning (SFT) as a preliminary step. The result? Outstanding reasoning capabilities with minimal labeled data
DeepSeek-R1-Zero is trained with almost pure RL, without supervised fine-tuning (SFT) as a preliminary step. The result? Outstanding reasoning capabilities with minimal labeled data