GraspMolmo: Generalizable Task-Oriented Grasping via Large-Scale Synthetic Data Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Deshpande, Abhay, Deng, Yuquan, Ray, Arijit, Salvador, Jordi, Han, Winson, Duan, Jiafei, Zeng, Kuo-Hao, Zhu, Yuke, Krishna, Ranjay, Hendrix, Rose
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908536523456512
author Deshpande, Abhay
Deng, Yuquan
Ray, Arijit
Salvador, Jordi
Han, Winson
Duan, Jiafei
Zeng, Kuo-Hao
Zhu, Yuke
Krishna, Ranjay
Hendrix, Rose
author_facet Deshpande, Abhay
Deng, Yuquan
Ray, Arijit
Salvador, Jordi
Han, Winson
Duan, Jiafei
Zeng, Kuo-Hao
Zhu, Yuke
Krishna, Ranjay
Hendrix, Rose
contents We present GrasMolmo, a generalizable open-vocabulary task-oriented grasping (TOG) model. GraspMolmo predicts semantically appropriate, stable grasps conditioned on a natural language instruction and a single RGB-D frame. For instance, given "pour me some tea", GraspMolmo selects a grasp on a teapot handle rather than its body. Unlike prior TOG methods, which are limited by small datasets, simplistic language, and uncluttered scenes, GraspMolmo learns from PRISM, a novel large-scale synthetic dataset of 379k samples featuring cluttered environments and diverse, realistic task descriptions. We fine-tune the Molmo visual-language model on this data, enabling GraspMolmo to generalize to novel open-vocabulary instructions and objects. In challenging real-world evaluations, GraspMolmo achieves state-of-the-art results, with a 70% prediction success on complex tasks, compared to the 35% achieved by the next best alternative. GraspMolmo also successfully demonstrates the ability to predict semantically correct bimanual grasps zero-shot. We release our synthetic dataset, code, model, and benchmarks to accelerate research in task-semantic robotic manipulation, which, along with videos, are available at https://abhaybd.github.io/GraspMolmo/.
format Preprint
id arxiv_https___arxiv_org_abs_2505_13441
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle GraspMolmo: Generalizable Task-Oriented Grasping via Large-Scale Synthetic Data Generation
Deshpande, Abhay
Deng, Yuquan
Ray, Arijit
Salvador, Jordi
Han, Winson
Duan, Jiafei
Zeng, Kuo-Hao
Zhu, Yuke
Krishna, Ranjay
Hendrix, Rose
Robotics
We present GrasMolmo, a generalizable open-vocabulary task-oriented grasping (TOG) model. GraspMolmo predicts semantically appropriate, stable grasps conditioned on a natural language instruction and a single RGB-D frame. For instance, given "pour me some tea", GraspMolmo selects a grasp on a teapot handle rather than its body. Unlike prior TOG methods, which are limited by small datasets, simplistic language, and uncluttered scenes, GraspMolmo learns from PRISM, a novel large-scale synthetic dataset of 379k samples featuring cluttered environments and diverse, realistic task descriptions. We fine-tune the Molmo visual-language model on this data, enabling GraspMolmo to generalize to novel open-vocabulary instructions and objects. In challenging real-world evaluations, GraspMolmo achieves state-of-the-art results, with a 70% prediction success on complex tasks, compared to the 35% achieved by the next best alternative. GraspMolmo also successfully demonstrates the ability to predict semantically correct bimanual grasps zero-shot. We release our synthetic dataset, code, model, and benchmarks to accelerate research in task-semantic robotic manipulation, which, along with videos, are available at https://abhaybd.github.io/GraspMolmo/.
title GraspMolmo: Generalizable Task-Oriented Grasping via Large-Scale Synthetic Data Generation
topic Robotics
url https://arxiv.org/abs/2505.13441