Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Unifying Deep Predicate Invention with Pre-trained Foundation Models

About

Long-horizon robotic tasks are hard due to continuous 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 language 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 times higher success rates than top-down methods and 3-4 times faster learning than bottom-up approaches, advancing scalable and flexible symbolic world modeling for robotics.

Qianwei Wang, Bowen Li, Zhanpeng Luo, Yifan Xu, Alexander Gray, Tom Silver, Sebastian Scherer, Katia Sycara, Yaqi Xie• 2025

Related benchmarks

TaskDatasetResultRank
Task PlanningSatellites SE2 (train)
Success Rate99.6
9
Task PlanningSatellites SE2 (test)
Success Rate95.2
9
Task PlanningBlocks Vec3 distribution (train)
Success Rate100
9
Task PlanningBlocks Vec3 (test)
Success Rate81.6
9
Task PlanningTable Clean Sim SE2 distribution (train)
Success Rate96.4
9
Task PlanningTable Clean Sim SE2 (test)
Success Rate93.4
9
Task PlanningTools PCD (train)
Success Rate100
8
Task PlanningTools PCD distribution (test)
Success Rate100
8
Task PlanningPacking Image (train)
Success Rate1
8
Task PlanningPacking Image (test)
Success Rate100
8
Showing 10 of 12 rows

Other info

Follow for update