Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling
논문 링크 Janus-Pro 7B: Dual-Encoder Multimodal LLM That Outsmarts Bigger Models한 줄 요약 (TL;DR)SigLIP 이해 인코더 + VQ 생성 인코더를 완전히 분리한 뒤 7 B 파라미터 LLM에 …
34분
DeepSeek
Dual Encoder
Multimodal Learning
Vision-Language Models
Large Language Models
Efficient Training