1
Mags-RL: Wearing Multimodal LLMs a Magnifying Glass via Agentic Reinforcement Learning For Complex Scene Reasoning
当多模态大模型戴上“放大镜”：新框架用智能体强化学习攻克复杂场景推理瓶颈，突破高密度遮挡与背景干扰。
arXiv:2605.27960v1 Announce Type: new Abstract: Despite their popularity and success, Multimodal Large Language Models (MLLMs) often struggle to inter…