Newest papers

PhysiCalWorld: In-Context Physics Calibration for World Models

Yunhao Luo, He Wentao, Yinpei Dai, Hyelin Nam, Hongze Fu, Joyce Yue Chai, Nima Fazeli, Jeong Joon Park · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.30125 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Os modelos mundiais existentes geralmente percebem uma cena através de um quadro visual atual ou de uma janela de história estreita. No entanto, essas observações podem não revelar as propriedades físicas que determinam a dinâmica de transição, como a distribuição de massa de um objeto, o amortecimento articulado ou a fricção da superfície. Sem calibrar esses fatores físicos para modelos mundiais, as previsões tendem a regressar em direção a uma antecedência empírica e induzir um desvio severo em cenas de teste novas. Para este fim, propomos PhysiCalWorld, que calibra os modelos mundiais condicionados por ação para a dinâmica de instância específica por meio de prompting em contexto, sem a modelação de física analítica. Nosso método codifica um clipe de interação

Machine translation (Meta M2M100); the English original follows.

Abstract

Existing world models typically perceive a scene through current visual frame or a narrow history window. However, these observations may not reveal the physics properties that determine transition dynamics, such as an object's mass distribution, joint damping, or surface friction. Without calibrating these physics factors for world models, predictions tend to regress toward an empirical prior and induce severe drift in novel test scenes. To this end, we propose PhysiCalWorld, which calibrates action-conditioned world models to the instance-specific dynamics via in-context prompting, without fine-tuning or analytical physics modeling. Our method encodes an exploratory interaction clip collected within the test scene into a compact physics descriptor token, which grounds subsequent predictions in the scene's actual physical behavior. PhysiCalWorld is trained end-to-end with the latent prediction loss alone, without physics labels. At inference time, the descriptor token is extracted once and can be reused across rollouts with negligible compute overhead. We evaluate PhysiCalWorld across four environments covering diverse physics variations and show that it substantially improves rollout prediction accuracy and visual planning success rate. The learned descriptors also form a physically structured latent space and generalize beyond the physics ranges seen during training.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Inference, Calibration, Physics beyond the Standard Model, Physics, Computer science, Space (punctuation)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange