研究人员正在开发新的框架和基准来增强大型语言模型（LLMs）的多模态推理能力。CoWorld Reasoner 引入了一个代理框架，该框架使用视频世界模型进行灵活的视觉想象和推理，在直接推理方面有所改进。ReasonUni 提出了一种推理增强的重建后训练方法，以耦合图像到文本的理解和文本到图像的生成。MagiC 提供了一个评估基于视觉的多模态认知的基准，评估推理的有效性以及与视觉证据的一致性，揭示了当前模型的局限性。L2T 提供了一个统一的框架，该框架联合优化视觉感知和文本推理，与显式工具调用方法相比，实现了更高的吞吐量。最后，Humanity's Sixth Sense 基准强调，当前前沿的多模态大型语言模型（MLLMs）在直观视觉推理任务上的表现远不如人类。
AI
arXiv:2610.21910v1 Announce Type: new Abstract: As Vision-Language Models (VLMs) advance, they are increasingly expected to perform spatiotemporal reasoning over motion, actions, and view changes beyond what the input images show. Video World Models (VWMs) provide a natural unifi…
arXiv cs.AI
TIER_1English(EN)·Feize Wu, Zhiyuan Yan, Yuhuan Hong, Yuwei Niu, Zongjian Li, Zheyuan Liu, Bin Lin, Jinbo Yan, Jie Qin, Xudong Mao, Limeng Qiao, Guanglu Wan, Li Yuan·
arXiv:2507.07297v2 Announce Type: cross Abstract: Recent advances in large vision-language models have led to impressive performance in visual question answering and multimodal reasoning. However, it remains unclear whether these models genuinely perform grounded visual reasoning…
arXiv cs.AI
TIER_1English(EN)·Quanlin Chen, Xiuwei Chen, Kun Xiang, Zehua Ma, Chunjin Ye, Boqiang Guo, Xianming Liu, Xiaodan Liang·
arXiv:2610.15810v1 Announce Type: cross Abstract: The introduction of long chain-of-thought (CoT) reasoning and external tools has substantially improved the performance of multimodal large language models (MLLMs) on general and reasoning tasks. However, existing fast-slow reason…
CoWorld Reasoner is an agentic framework that uses a video world model as a flexible visual reasoning engine, enabling vision-language models to perform diverse spatiotemporal reasoning through guided visual imagination and co-evolving tools.
Humans perceive far more in a scene than what is explicitly depicted: a single glance captures past causes and future trajectories; a quick peek determines if a vehicle can fit between two parked cars; a few seconds of video reveals who holds authority in a room; and a fleeting c…
arXiv cs.CV
TIER_1English(EN)·Yiming Zhang, Qiangyu Yan, Borui Jiang, Kai Han·
arXiv:2602.05359v3 Announce Type: replace Abstract: The advancement of multimodal large language models (MLLMs) has enabled impressive perception capabilities. However, their reasoning process often remains a "fast thinking" paradigm, reliant on end-to-end generation or explicit,…
