Sungdong Kim
@sungdongkim4
Research Scientist @ NAVER Cloud; MS&PhD student @ KAIST #NLP #LLM #Alignment
ID: 1390479880698007555
07-05-2021 01:33:40
92 Tweet
419 Takipçi
183 Takip Edilen
@sungdongkim4
Research Scientist @ NAVER Cloud; MS&PhD student @ KAIST #NLP #LLM #Alignment
ID: 1390479880698007555
07-05-2021 01:33:40
92 Tweet
419 Takipçi
183 Takip Edilen
![🤔 Do we always need a human preference for effective LLM alignment after an SFT stage? Our answer is NO 🙅♂️
We present a ✨preference-free alignment approach✨, leveraging an off-the-shelf retriever with effective regularizer functions: Regularized Relevance Reward (R^3). [1/n]](https://pbs.twimg.com/media/GF5UuoZbgAAjzFo.jpg "Sungdong Kim (@sungdongkim4) on Twitter photo 🤔 Do we always need a human preference for effective LLM alignment after an SFT stage? Our answer is NO 🙅♂️
We present a ✨preference-free alignment approach✨, leveraging an off-the-shelf retriever with effective regularizer functions: Regularized Relevance Reward (R^3). [1/n]")