PulseAugur
EN
LIVE 18:02:07

New models FloMo and WorldFlow predict 3D motion from video

By PulseAugur Editorial · [2 sources] ·

Researchers have developed FloMo and WorldFlow, two distinct models focused on predicting 3D scene motion and action from video. FloMo, a 5 billion parameter model, fine-tunes video generation capabilities to predict 3D scene flow and action tokens, demonstrating that motion prediction is more effective for robot action generalization than direct video prediction. WorldFlow, utilizing a Mixture-of-Transformers backbone, jointly predicts future video and 3D trajectories using continuous displacement tokens, showing improved physical dynamics prediction and significantly reducing error on benchmarks like SpatialCog. AI

IMPACT These models advance the ability of AI to understand and predict physical dynamics from visual input, crucial for robotics and embodied AI.

RANK_REASON Two distinct research papers introducing new models for 3D motion prediction from video.

Read on Hugging Face Daily Papers →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

New models FloMo and WorldFlow predict 3D motion from video

How we ranked this

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Research
Two distinct research papers introducing new models for 3D motion prediction from video.
Source corroboration
2 independent sources
Multiple independent publishers reporting the same story raises confidence that it's real and newsworthy.
Topics
model release, paper
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
15 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

Full methodology in our editorial standards.

COVERAGE [2]

  1. arXiv cs.AI TIER_1 English(EN) · Jeremy Collins, Namra Patel, Ayush Agarwal, Shitij Govil, Animesh Garg ·

    FloMo: 3D Scene Flow as a World Action Model Intermediate for Learning from Human Video

    arXiv:2611.05062v2 Announce Type: replace-cross Abstract: Pretrained video generation models distill internet-scale motion priors into robot policies, but full video is a redundant signal for action prediction. We argue that dense 3D motion is the medium through which video pretr…

  2. Hugging Face Daily Papers TIER_1 English(EN) ·

    WorldFlow: Joint Video and 3D Motion Prediction via Large-Scale Pretraining

    WorldFlow is a generative world model that jointly predicts future video and 3D trajectories using continuous displacement tokens and a Mixture-of-Transformers backbone, improving physical dynamics prediction across diverse benchmarks.

RELATED ENTITIES

RELATED TOPICS