# Mags-RL: Wearing Multimodal LLMs a Magnifying Glass via Agentic Reinforcement Learning For Complex Scene Reasoning Source: [Mags-RL: Wearing Multimodal LLMs a Magnifying Glass via Agentic Reinforcement Learning For Complex Scene Reasoning]() Collection index: [llms.txt]() Data updated: 2026-05-29T05:00:05.519Z Authors: Xuanzhao Dong, Wenhui Zhu, Peijie Qiu, Xiwen Chen, Xiaobing Yu, Xin Li, Zhipeng Wang, Shao Tang, Gen Li, Yujian Xiong, Hao Wang, Yanxi Chen, Prayag Tiwari, Yalin Wang Published: 2026-05-27T04:54:07.000Z ## Abstract Despite their popularity and success, Multimodal Large Language Models (MLLMs) often struggle to interpret images accurately, which limits their reasoning capability in complex scenarios (e.g., high object density and complex background clutter). Prior work mainly addresses this limitation by incorporating explicit visual cues like bounding boxes that require extra annotations. In addition, the resulting low-resolution crops often miss fine-grained details that MLLMs require for accurate reasoning. Therefore, we propose Mags-RL, an Agentic Reinforcement Learning (RL) framework that equips MLLMs with an external super-resolution "magnifying glass" agent for high-resolution fine-grained inspection. Specifically, the model performs two-round reasoning: in the first round, it generates an initial rationale and autonomously identifies regions of interest without relying on additional annotations; in the second round, it invokes a super-resolution agent to crop and upscale those regions, then revisits and verifies its earlier reasoning to produce the final answer. We also introduce a novel curriculum learning strategy that enables data-efficient RL training, needing as few as only 40 training samples to achieve reasonable performance. Experiments on VSR, TallyQA, and GQA subsets show its superior performance against recent strong competing methods, demonstrating high-quality reasoning with precise visual grounding. Code and weights will be released soon. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Visual Question Answering on TallyQA ZOOM-Easy](): Visual Question Answering; TallyQA ZOOM-Easy; GPT Accuracy: 51.27. - [Visual Question Answering on GQA ZOOM-Easy](): Visual Question Answering; GQA ZOOM-Easy; GPT Accuracy: 68.75. - [Visual Question Answering on TallyQA ZOOM-Medium](): Visual Question Answering; TallyQA ZOOM-Medium; GPT Accuracy: 24.35. - [Visual Question Answering on GQA ZOOM-Medium](): Visual Question Answering; GQA ZOOM-Medium; GPT Accuracy: 59.31. - [Visual Question Answering on TallyQA (ZOOM-Hard)](): Visual Question Answering; TallyQA (ZOOM-Hard); GPT Accuracy: 16.78. - [Visual Question Answering on GQA ZOOM-Hard](): Visual Question Answering; GQA ZOOM-Hard; GPT Accuracy: 51.44. - [Visual Spatial Reasoning on VSR (ZOOM-Easy)](): Visual Spatial Reasoning; VSR (ZOOM-Easy); GPT Accuracy: 73.09. - [Visual Spatial Reasoning on VSR ZOOM-Medium](): Visual Spatial Reasoning; VSR ZOOM-Medium; GPT Accuracy: 67.63. - [Visual Spatial Reasoning on VSR ZOOM-Hard](): Visual Spatial Reasoning; VSR ZOOM-Hard; GPT Accuracy: 55.98. ## Original sources and implementation [Original paper]() [PDF]()