IPR-1：交互式物理推理器
IPR-1: Interactive Physical Reasoner
用世界模型 rollout 作为奖励信号来强化 VLM 策略，并用物理中心动作码（PhysCode）对齐语义意图与动力学，为视觉研究者提供了一种在交互环境中把预测模型与推理模型耦合的可复用范式。
原文依据 using world-model rollouts to score and reinforce a VLM's policy
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论人类通过观察、与环境交互并内化物理与因果关系来学习。我们探索智能体能否同样通过交互与经验获得类人的推理能力。为此，我们提出了 Game-to-Unseen（G2U）基准，包含 1000 多个存在显著视觉域差异的异构游戏。现有方法（包括视觉语言模型与世界模型）由于对视觉细节过拟合，难以把握底层物理与因果关系。VLM/VLA 智能体能够推理，但在交互场景中缺乏前瞻能力；世界模型能够想象，但只是模仿视觉模式，而非分析物理与因果关系。因此我们提出 IPR（交互式物理推理器），利用世界模型的 rollout 对 VLM 的策略进行打分与强化，并引入 PhysCode——一种以物理为中心的动作码，将语义意图与动力学对齐，为预测与推理提供共享的动作空间。IPR 在 1000 多个游戏上预训练，在从原始直觉到目标驱动推理的各层级上均表现稳健，甚至总体超越 GPT-5。我们发现性能随训练游戏数量与交互步数的增加而提升，且模型能够零样本迁移到未见过的游戏。这些结果支持以物理为中心的交互作为持续提升物理推理能力的路径。更多演示与细节见 https://mybearyzhang.github.io/ipr-1。
英文原文摘要 1393 字符
Humans learn by observing, interacting with environments, and internalizing physics and causality. We explore whether agents can similarly acquire human-like reasoning through interaction and experience. To study this, we introduce a Game-to-Unseen (G2U) benchmark of 1,000+ heterogeneous games that exhibit significant visual domain gaps. Existing approaches, including VLMs and world models, struggle with underlying physics and causality due to overfitting on visual details. VLM/VLA agents reason but lack look-ahead in interactive settings, while world models imagine but imitate visual patterns rather than analyze physics and causality. We therefore propose IPR (Interactive Physical Reasoner), using world-model rollouts to score and reinforce a VLM's policy, and introduce PhysCode, a physics-centric action code aligning semantic intent with dynamics to provide a shared action space for prediction and reasoning. Pretrained on 1,000+ games, IPR performs robustly on levels from primitive intuition to goal-driven reasoning, and even surpasses GPT-5 overall. We find that performance improves with more training games and interaction steps, and that the model also zero-shot transfers to unseen games. These results support physics-centric interaction as a path to steadily improving physical reasoning. Further demos and details can be found at https://mybearyzhang.github.io/ipr-1.