
J
@long_text_
Followers
30
Following
932
Media
9
Statuses
585
Joined October 2020
RT @Muy_bien_Y_tu_9: ��재 전지구상에 총 24억대의 에어컨이 존재하지만 향후 25년 안에 에어컨은 50억대 이상으로 늘어날 전망. 과연 그때까지 우리는 에너지 문제를 해결할수있을런지.
0
15
0
RT @jxmnop: OpenAI hasn’t open-sourced a base model since GPT-2 in 2019. they recently released GPT-OSS, which is reasoning-only. or is….
0
457
0
RT @burkov: When it’s a word pattern matcher, it’s a word pattern matcher. You might think it’s intelligence, but it’s a word pattern match….
0
500
0
RT @MichaelAArouet: Important chart. S&P 490 has had basically no earnings growth since 2022, despite rampant inflation. It’s just 10 compa….
0
3K
0
일견 흥미로워보일 수 있는 현상이지만 사실은 원래 저렇게 대답하도록 된 cloud 모델을 distill 했기 때문일 것. 그런데 저런 종류의 응답을 RLHF로 트레이닝 맞는 걸까? 스스로 찾고 생각해서 대답해야 할 것을 외워버리게 하는 것인���.
one pattern i’ve noticed is that open weights models from big us labs get very defensive and disbelieving if you tell the assistant persona it’s an open-weights model. also happens with gemma
0
0
0