Goal Force: Teaching Video Models To Accomplish Physics-Conditioned Goals

TL;DR

This paper introduces Goal Force, a framework for teaching video models to understand and generate physics-conditioned goals using explicit force vectors, enabling better physical reasoning and planning in complex scenarios.

Contribution

The paper proposes a novel force-based goal specification method and trains a video model on synthetic physics data, demonstrating zero-shot generalization to real-world physics tasks.

Findings

Model generalizes to complex real-world scenarios

Grounding in physical interactions enables physics-aware planning

Zero-shot transfer from synthetic to real-world physics tasks

Abstract

Recent advancements in video generation have enabled the development of ``world models'' capable of simulating potential futures for robotics and planning. However, specifying precise goals for these models remains a challenge; text instructions are often too abstract to capture physical nuances, while target images are frequently infeasible to specify for dynamic tasks. To address this, we introduce Goal Force, a novel framework that allows users to define goals via explicit force vectors and intermediate dynamics, mirroring how humans conceptualize physical tasks. We train a video generation model on a curated dataset of synthetic causal primitives-such as elastic collisions and falling dominos-teaching it to propagate forces through time and space. Despite being trained on simple physics data, our model exhibits remarkable zero-shot generalization to complex, real-world scenarios, including tool manipulation and multi-object causal chains. Our results suggest that by grounding video generation in fundamental physical interactions, models can emerge as implicit neural physics simulators, enabling precise, physics-aware planning without reliance on external engines. We release all datasets, code, model weights, and interactive video demos at our project page.