MolmoMotion: Forecasting Point Trajectories in 3D with Language Instruction

2026-06-16

AI Summary

This work proposes the task of goal-conditioned 3D point motion prediction: given visual history, a natural language instruction, and a set of 3D query points on an object’s surface, the model predicts their future trajectories in 3D space. To support this task, we introduce MolmoMotion-1M, the first large-scale 3D point trajectory dataset with language annotations, along with PointMotionBench, a human-validated benchmark for evaluating motion prediction. We further design a language-guided, general-purpose motion prediction framework that combines autoregressive coordinate prediction with flow-matching-based trajectory generation, trained end-to-end in world coordinates. Experiments demonstrate that our model significantly outperforms existing approaches on PointMotionBench; the learned motion priors substantially enhance training efficiency and generalization in robotic manipulation and provide more realistic motion guidance for video generation.

Abstract

Motion forecasting is central to visual intelligence: agents must anticipate how objects will move in order to plan actions, reason about physical interactions, and synthesize realistic futures. We argue that 3D points in world coordinates provide a general representation that is class-agnostic, view-stable, compact, and directly useful for downstream tasks. We formalize the task of goal-conditioned 3D point motion forecasting: given a short visual history, a set of 3D query points on an object of interest, and a language description of the intended goal, the model predicts the future 3D trajectory of each point. We introduce a full stack to study this task at scale: (1) MolmoMotion-1M is a large corpus of action-described, object-grounded 3D point trajectories annotated from 1.16M unconstrained videos; (2) PointMotionBench is a human-verified benchmark spanning 111 object categories and 61 motion types; and (3) MolmoMotion is a general motion forecasting model that supports both autoregressive coordinate prediction and flow-matching-based trajectory generation. MolmoMotion accurately predicts diverse motion patterns with different language instructions, and significantly outperforms existing motion prediction baselines on PointMotionBench. Finally, we show that the learned 3D motion prior transfers well to downstream applications: it improves training efficiency and generalization for robot manipulation, and its predicted trajectories provide effective motion guidance for generative models to synthesize videos with more realistic object motion.

Problem

• 3D motion forecasting
• language instruction
• point trajectory prediction
• goal-conditioned motion
• visual intelligence

Innovation

• 3D motion forecasting
• language-conditioned prediction
• point trajectory
• flow matching
• motion prior