Paper review29 Prompt-prompted Mixture of Experts for Efficient LLM Generation 논문 리뷰 오늘 리뷰해 볼 논문은 제가 평소에 LInkdin에서 평소 논문 LLM , RAG 관련 리서치할 때 종종 보는 Pascal Biese 가 소개해준 'Prompt-prompted Mixture of Experts for Efficient LLM Generation '이라는 논문입니다. 궁금해서 공유하기를 눌러두고 오늘 리뷰를 해봅니다. Paper Prompt-prompted Mixture of Experts for Efficient LLM Generation With the development of transformer-based large language models (LLMs), they have been applied to many fields due to their remarkable utili.. 2024. 4. 15. Gecko: Versatile Text Embeddings Distilledfrom Large Language Models 한글 리뷰 Gecko: Versatile Text Embeddings Distilled from Large Language Models We present Gecko, a compact and versatile text embedding model. Gecko achieves strong retrieval performance by leveraging a key idea: distilling knowledge from large language models (LLMs) into a retriever. Our two-step distillation process begins with gen arxiv.org Abstract Gecko는 큰 언어 모델(Large Language Models, LLMs)에서 지식을 추출.. 2024. 4. 2. Jamba:A Hybrid Transformer-Mamba Language Model 리뷰 Jamba: A Hybrid Transformer-Mamba Language Model We present Jamba, a new base large language model based on a novel hybrid Transformer-Mamba mixture-of-experts (MoE) architecture. Specifically, Jamba interleaves blocks of Transformer and Mamba layers, enjoying the benefits of both model families. MoE is arxiv.org ai21labs/Jamba-v0.1 · Hugging Face Model Card for Jamba Jamba is a state-of-the-art.. 2024. 4. 2. sDPO: Don’t Use Your Data All at Once 한글 리뷰 Upstage에서 발표한 논문입니다. 기존의 DPO 방법으로 SOLAR를 학습시켜서 현재 PLM으로 Langchain에서도 쓸 수 있는 모델이 되었습니다. gram 노트북에도 들어가고 폴라리스 랩소디에서 Copilot 형태처럼 solar를 탑재 등 다양한 뉴스가 나오고 있는 토종의 모델입니다. DPO 학습을 누구보다 잘 알기에 새로운 방법론 sDPO에 대한 논문이 나온 것 같습니다. https://arxiv.org/abs/2403.19270 sDPO: Don't Use Your Data All at Once As development of large language models (LLM) progresses, aligning them with human preferences has become incre.. 2024. 4. 1. 이전 1 2 3 4 5 ··· 8 다음
