[논문리뷰] Where-OPD: Spatially Guided On-Policy Self-Distillation of MLLMs with Synthetic Scenes본 논문은 MLLMs의 시각적 추론 능력을 향상시키기 위한 새로운 On-policy self-distillation 접근법을 제안합니다.#Review#Multimodal Large Language Models#On-Policy Self-Distillation#Synthetic Data#Visual Reasoning#Spatially Grounded Guidance#Post-Training#Synthetic-to-Real Transfer2026년 10월 1일댓글 수 로딩 중