询问世界：通过智能体世界建模与探测实现通用物理推理
Asking the World: Generalist Physical Reasoning through Agentic World Modeling and Probing
- Tsinghua University（清华大学）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
提出ATW智能体，通过世界建模与探测构建可执行世界进行物理推理，在多个基准上显著超越现有方法。
AI中文摘要：
从视频中进行物理推理需要推断超出直接观察范围的潜在物理属性和动态。在没有显式建模和验证的情况下，直接使用视觉语言模型（VLM）进行推理在复杂物理任务上仍然不可靠，而预定义的工具流水线依赖于任务和领域特定的先验知识，这限制了其在材料、动态和推理任务上的泛化能力。我们提出了“询问世界”（Asking the World, ATW），一个通过两个自适应阶段构建并询问与任务相关的可执行世界的通用智能体：世界建模（World Modeling）从视频中校准一个世界，而世界探测（World Probing）则对该世界进行查询、模拟和干预，以获取与问题相关的证据。ATW 并不预设这两个阶段中的操作，而是根据场景和问题决定如何建模和探测。我们开发了 PolyWorld Engine，一个基于 Warp 的轻量级且高度可编程的多物理场模拟器，用于构建和探测包含刚体、软体、布料、绳索、流体及其耦合交互的世界。基于交叉熵方法（CEM）的系统辨识在世界建模过程中恢复与任务相关的动态。由此产生的世界成为一个用于问题导向物理实验的主动工作空间，而非预定义的下游工具。我们在 CLEVRER、ContPhy 和三个真实世界场景上评估了 ATW。使用 Gemini-3-Flash 作为其基础视觉语言模型，ATW 在 CLEVRER 上达到了 80.82% 的总体每问题准确率，比直接使用 Gemini-3-Flash 提高了 46.50 个百分点，比 GPT-5.5 提高了 13.58 个百分点，比 PhysMind 提高了 8.27 个百分点。在 ContPhy 上，它达到了 70.56% 的总体准确率，比 Gemini-3-Flash 提高了 28.10 个百分点，比 GPT-5.5 提高了 3.53 个百分点。在三个真实世界场景中，ATW 达到了 71.67% 的准确率，比 GPT-5.5 高出 28.33 个百分点。这些结果确立了智能体世界建模与探测作为一种有效的、基于执行的方法，用于实现通用物理推理。
英文摘要：
Physical reasoning from video requires inferring latent physical properties and dynamics beyond direct observation. Direct VLM inference remains unreliable on complex physical tasks without explicit modeling and validation, while predefined tool pipelines rely on task- and domain-specific priors that limit generalization across materials, dynamics, and reasoning tasks. We introduce Asking the World (ATW), a generalist agent that constructs and interrogates task-relevant executable worlds through two adaptive stages: World Modeling calibrates a world from video, while World Probing queries, simulates, and intervenes on it to obtain question-relevant evidence. Rather than prescribing the operations in either stage, ATW determines how to model and probe according to the scene and question. We develop PolyWorld Engine, a lightweight and highly programmable Warp-based multiphysics simulator for constructing and probing worlds with rigid bodies, soft bodies, cloth, ropes, fluids, and their coupled interactions. CEM-based system identification recovers task-relevant dynamics during World Modeling. The resulting world becomes an active workspace for question-directed physical experiments rather than a predetermined downstream tool. We evaluate ATW on CLEVRER, ContPhy, and three real-world scenarios. Using Gemini-3-Flash as its base VLM, ATW achieves 80.82% overall per-question accuracy on CLEVRER, improving direct Gemini-3-Flash by 46.50 points, GPT-5.5 by 13.58 points, and PhysMind by 8.27 points. On ContPhy, it reaches 70.56% overall accuracy, surpassing Gemini-3-Flash by 28.10 points and GPT-5.5 by 3.53 points. Across the three real-world scenarios, ATW achieves 71.67% accuracy, 28.33 points above GPT-5.5. These results establish agentic world modeling and probing as an effective, execution-grounded approach to generalist physical reasoning.
