SpatialSkill：跨视角空间推理的自进化技能

中文摘要 AI 辅助

跨视角空间推理要求模型将不同视角对齐为一致的空间表征，尽管这对人类而言是自然的，但这一能力对视觉-语言模型仍具挑战性。现有方法通常通过更新模型权重来改进空间推理，这使得所学知识保持隐式且与特定骨干网络绑定。我们提出SpatialSkill，一个免权重更新的框架，使冻结的视觉-语言模型能够从离线轨迹中积累显式的自然语言推理技能。与符号任务不同，感知技能无法仅通过执行来可靠验证：一个看似合理的空间规则可能缺乏视觉支持，或需要冻结模型无法执行的变换。因此，SpatialSkill仅在视觉锚定和可执行性检查后接纳候选技能，约束手动进化以防止有害回归，并按空间推理类别路由技能以减少负迁移。在CityCube上，跨四个冻结执行器，SpatialSkill带来一致的增益，配备SpatialSkill的9B执行器在我们评估中超过了最强的闭源参考。技能存储在带版本的自然语言手册中，使推理策略显式且可审计，无需修改模型参数。代码见https://this URL。

英文摘要

Cross-view spatial reasoning requires a model to align different viewpoints into a coherent spatial representation, yet this ability remains challenging for vision-language models despite being natural to humans. Existing methods typically improve spatial reasoning by updating model weights, which keeps the acquired knowledge implicit and tied to a specific backbone. We propose \textit{SpatialSkill}, a weight-update-free framework that enables a frozen vision-language model to accumulate explicit natural-language reasoning skills from offline trajectories. Unlike symbolic tasks, perceptual skills cannot be reliably verified simply by executing them: a plausible spatial rule may lack visual support or require transformations that the frozen model cannot perform. SpatialSkill therefore admits candidate skills only after visual-grounding and executability checks, constrains manual evolution to prevent harmful regressions, and routes skills by spatial-reasoning category to reduce negative transfer. On CityCube, across four frozen executors, SpatialSkill yields consistent gains, and a 9B executor equipped with SpatialSkill surpasses the strongest closed-source reference in our evaluation. The skills are stored in a versioned natural-language manual, making the reasoning strategies explicit and auditable without modifying model parameters. Code at https://github.com/vindahi/SpatialSkill.

发表机构

• Shandong University（山东大学）
• Shanghai AI Laboratory（上海人工智能实验室）
• Zhiyang Innovation Co., Ltd.（智洋创新科技股份有限公司）

机构由 AI 辅助整理，请以论文原文为准。