RLHF replaced with RLAIF , SFT used prior to RL techniques, More or less, would have used likes of LLAMA 3.* as the base SFT model and then implemented the RL technqiues. Ideal to treat it as just another so called "Reasoning Immitator" model. But yes, a free model. But performace difference across distilled models vs original models is exponential.
Yes, you are correct! DeepSeek-R1 follows RLAIF instead of RLHF, with SFT before RL techniques. And the performance gap between distilled and original models is quite significant!
Thank you
One of the best video on deepseek
Thanks!
Best video on DeepSeek
Thanks!
Very Nice video on deepseek
Thanks
wow mam you are to updated with new knowledge we want this type of teacher thanks
Thank you so much! 😊 I love exploring new technologies and sharing knowledge. Glad you find it helpful!
What a wonderful video on deepseek, its beautifully explained
Thanks!
Thank you for the latest info mam 🙂
Welcome 🙂
Thank you for your video 😊😊
Deepseek, amazing video
Thanks!
Great one
Thanks!
Helpful
Glad it helped
RLHF replaced with RLAIF , SFT used prior to RL techniques, More or less, would have used likes of LLAMA 3.* as the base SFT model and then implemented the RL technqiues. Ideal to treat it as just another so called "Reasoning Immitator" model. But yes, a free model. But performace difference across distilled models vs original models is exponential.
Yes, you are correct! DeepSeek-R1 follows RLAIF instead of RLHF, with SFT before RL techniques. And the performance gap between distilled and original models is quite significant!
Please learn us how to use it locally on our mobile phone
Noted!
Someone already made a video on this topic
any refral ?
Nice
Thanks
is android app available? safe to use? it's hakka noodles thing so...
Wow, I can really run it locally on my machine offline ?! 😮
Yes :)
Can you put a tutorial to build ai agent using deepseek phidata and gemini free llm
Noted!
I asked DeepSeek if it is free and open source. It said "I am a proprietary AI model developed by DeepSeek"
That's interesting!
Deep face detection application