笛卡尔捷径：在极坐标空间中重新评估视觉推理
The Cartesian Shortcut: Re-evaluate Vision Reasoning in Polar Coordinate Space
AI总结：
针对多模态大语言模型在视觉推理中利用笛卡尔坐标捷径的问题，提出Polaris-Bench基准，将任务转换至极坐标空间，揭示模型缺乏拓扑不变性视觉推理。
摘要：
随着当前多模态大语言模型迅速饱和标准视觉推理基准，一个关键问题浮现：这些高分是否真正反映了鲁棒的视觉理解？我们发现了一个普遍存在的漏洞，即笛卡尔捷径：视觉推理基准普遍基于正交网格布局，这些布局可以轻易地离散化为显式的文本坐标。模型系统地利用这一特性，大量依赖基于文本的演绎推理来辅助视觉问题解决。为了系统地消除这一捷径，我们引入了Polaris-Bench，该基准将53个视觉推理任务重新表述在极坐标空间中，并配有对应的笛卡尔坐标作为参考，同时保持一致的逻辑约束和任务语义——从而从根本上打破了模型所利用的正交先验。对14个最先进MLLM的全面评估显示，在笛卡尔布局上达到70%-83%的前沿模型在极坐标等价布局上骤降至31%-39%，即使在完全逻辑等价的情况下，性能下降依然持续。此外，在笛卡尔布局上观察到的推理增益在极坐标等价布局上严重减弱。这些发现揭示了当前MLLM的一个关键缺陷：缺乏拓扑不变的视觉推理。
英文摘要：
As current Multimodal Large Language Models rapidly saturate canonical visual reasoning benchmarks, a key question emerges: do these strong scores genuinely reflect robust visual understanding? We identify a pervasive vulnerability, the Cartesian Shortcut: models frequently discretize the orthogonal grid-based layouts prevalent in visual reasoning benchmarks into explicit textual coordinates, offloading reasoning from visual perception to text-based deduction. To re-evaluate visual reasoning when this shortcut is unavailable, we introduce Polaris-Bench, which re-formulates 53 visual reasoning tasks in Polar coordinate space with paired Cartesian counterparts that preserve task semantics, disrupting the orthogonal structure that models exploit. Comprehensive evaluation across $14$ state-of-the-art MLLMs reveals that frontier models achieving $69$--$83\%$ on Cartesian layouts collapse to $31$--$39\%$ on Polar equivalents. Moreover, thinking gains largely vanish on Polar layouts, prompting interventions fail to close the gap, and comparable drops arise on other non-orthogonal layouts. These findings reveal that current MLLMs' visual reasoning performance is strongly coupled to orthogonal grid structure, a fragility that is consistent across model families but far smaller in humans, who maintain 88.8\% accuracy on Polar layouts. The benchmark, evaluation suite, and leaderboard are publicly available at https://google-deepmind.github.io/polaris-bench.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
