# ExoPredicator: Learning Abstract Models of Dynamic Worlds for Robot Planning

Source: [ExoPredicator: Learning Abstract Models of Dynamic Worlds for Robot Planning](<https://www.sota2.com/research/paper/68fa7c321e209131cea8cf4c>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-18T05:00:11.457Z

Authors: Yichao Liang, Dat Nguyen, Cambridge Yang, Tianyang Li, Joshua B. Tenenbaum, Carl Edward Rasmussen, Adrian Weller, Zenna Tavares, Tom Silver, Kevin Ellis

Published: 2025-09-30T13:44:34.000Z

## Abstract

Long-horizon embodied planning is challenging because the world does not only change through an agent's actions: exogenous processes (e.g., water heating, dominoes cascading) unfold concurrently with the agent's actions. We propose a framework for abstract world models that jointly learns (i) symbolic state representations and (ii) causal processes for both endogenous actions and exogenous mechanisms. Each causal process models the time course of a stochastic cause-effect relation. We learn these world models from limited data via variational Bayesian inference combined with LLM proposals. Across five simulated tabletop robotics environments, the learned models enable fast planning that generalizes to held-out tasks with more objects and more complex goals, outperforming a range of baselines.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Robotic Planning on Grow](<https://www.sota2.com/research/sota/robotic-planning-on-grow.md>): Robotic Planning; Grow; Success Rate: 93.3.

- [Robotic Planning on Domino](<https://www.sota2.com/research/sota/robotic-planning-on-domino.md>): Robotic Planning; Domino; Success Rate: 98.7.

- [Robotic Planning on Coffee](<https://www.sota2.com/research/sota/robotic-planning-on-coffee.md>): Robotic Planning; Coffee; Success Rate: 99.3.

- [Robotic Planning on Boil](<https://www.sota2.com/research/sota/robotic-planning-on-boil.md>): Robotic Planning; Boil; Success Rate: 92.7.

- [Robotic Planning on Fan](<https://www.sota2.com/research/sota/robotic-planning-on-fan.md>): Robotic Planning; Fan; Success Rate: 97.3.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2509.26255>)

[PDF](<https://arxiv.org/pdf/2509.26255.pdf>)
