arXiv cs.CL· Shoubin Yu, Yue Zhang, Zun Wang, Jaehong Yoon, Huaxiu Yao, Mingyu Ding, Mohit Bansal·· 10 小时前AI 评分32
AVIC:用世界模型自适应控制测试时视觉想象,提升多模态模型空间推理
When and How Much to Imagine: Adaptive Test-Time Scaling with World Models for Visual Spatial Reasoning
AI 导读
研究提出 AVIC 自适应测试时框架,先判断当前视觉证据是否充分,再选择性调用并缩放世界模型进行视觉想象。其 AVIC-R 版本通过 GRPO 以答案正确性奖励和想象成本惩罚训练该决策策略,无需标注何时该想象、该想象多少。在 SAT、MMSI 空间推理基准和 R2R 具身导航基准上,AVIC-R 超过 GPT-4o 和 GPT-4.1 等闭源基线,且调用世界模型次数更少。
正文 · 原文
Abstract:Despite rapid progress in MLLMs, visual spatial reasoning remains unreliable when correct answers depend on how a scene would appear under unseen or alternative viewpoints. Recent work addresses this by augmenting reasoning with world models for visual imagination, but questions such as when imagination is actually necessary, how much of it is beneficial, and when it becomes harmful, remain poorly understood. In practice, indiscriminate imagination can increase computation and even degrade performance by introducing misleading evidence. In this work, we present an in-depth analysis of test-time visual imagination as a controllable resource for spatial reasoning. We first study when static visual evidence is sufficient, when imagination improves reasoning, and how excessive or unnecessary imagination affects accuracy and efficiency. To support this analysis, we then introduce AVIC, an adaptive test-time framework with world models that explicitly reasons about the sufficiency of current visual evidence before selectively invoking and scaling visual imagination. Finally, to further learn this gating and planning behavior without any annotation of when and how much to imagine, we introduce AVIC-R, which trains the policy via GRPO from QA-correctness rewards and penalties by imagination cost. Across spatial reasoning benchmarks (SAT, MMSI) and an embodied navigation benchmark (R2R), our results reveal clear scenarios where imagination is critical, marginal, or detrimental, and show that selective control can match or outperform fixed imagination strategies with substantially fewer world-model calls and language tokens. Our AVIC-R surpasses strong proprietary baselines including GPT-4o and GPT-4.1 while invoking the world model less often. Overall, our findings highlight the importance of analyzing and controlling test-time imagination for efficient and reliable spatial reasoning.
| Comments: | NeurIPS 2026; the first two authors are equally contributed. Project page: this https URL |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2602.08236 [cs.CV] |
| (or arXiv:2602.08236v3 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2602.08236 arXiv-issued DOI via DataCite |
Submission history
From: Shoubin Yu [view email]
[v1]
Mon, 9 Feb 2026 03:21:48 UTC (2,315 KB)
[v2]
Sun, 31 May 2026 23:44:59 UTC (2,438 KB)
[v3]
Wed, 18 Nov 2026 22:05:06 UTC (2,442 KB)
来源:arXiv cs.CL · arxiv.org