∗ Equal Contribution. † Corresponding author.
Contact: [email protected].
The work was partly done when Qianwei Wang and Zhanpeng Luo were Robotics Institute Summer Scholars associated with the Advanced Agent Robotics Technology Lab, CMU.
Long horizon robotic tasks are hard due to contin uous state action spaces and sparse feedback. Symbolic world models help by decomposing tasks into discrete predicates that capture object properties and relations. Existing methods learn predicates either top down, by prompting foundation models without data grounding, or bottom up, from demonstrations without high level priors. We introduce UniPred, a bilevel learning framework that unifies both. UniPred uses large lan guage models (LLMs) to propose predicate effect distributions that supervise neural predicate learning from low level data, while learned feedback iteratively refines the LLM hypotheses. Leveraging strong visual foundation model features, UniPred learns robust predicate classifiers in cluttered scenes. We further propose a predicate evaluation method that supports symbolic models beyond STRIPS assumptions. Across five simulated and one real robot domains, UniPred achieves 2 ∼ 4× higher success rates than top down methods and 3 ∼ 4× faster learning than bottom up approaches, advancing scalable and flexible symbolic world modeling for robotics.
UniPred is a novel framework that integrates deep learning and symbolic reasoning for long-horizon planning tasks in robotics. By leveraging large language models (LLMs) and visual foundation models, UniPred can effectively learn and reason about complex object properties and relationships in cluttered environments.
Figure 1: Overview of UniPred framework.
Figure 2: Overview of unified bilevel learning.
Caption for main demo 1.
Caption for main demo 2.
UniPred continuously replans based on observation feedback, detects action failures during execution, and performs recovery replanning. Through a total of 13 steps, the system successfully completes the task.
Long horizon manipulation tasks are inherently challenging. Many unpredictable factors can lead to failure, including planning failures and accumulated execution errors at each action step.
@misc
{
wang2025unifyingdeeppredicateinvention,
title={Unifying Deep Predicate Invention with Pre-trained Foundation Models},
author={Qianwei Wang and Bowen Li and Zhanpeng Luo and Yifan Xu and Alexander Gray and Tom Silver and Sebastian Scherer and Katia Sycara and Yaqi Xie},
year={2025},
eprint={2512.17992},
archivePrefix={arXiv},
primaryClass={cs.RO},
url={https://arxiv.org/abs/2512.17992},
}