cs.AINov 9, 2026

Learning to Act with World Models by Valuing What Comes Next

Authors: Chenghao Gao, Zhi Zhang, Dian Jin, Yihe Pan, Doudou Zhou

Organizations: Department of Statistics and Data Science, National University of Singapore · Department of Statistics and Data Science, University of California, Los Angeles · Department of Data Science, Fudan University

Abstract

A sparse terminal reward tells an agent little about its steps. Much of the missing information is implicit in the environment's dynamics, because each state transition can change the conditions for success long before the reward arrives. A world model trained on real transitions can learn to predict what each action changes, but a predicted observation does not by itself show whether success has become more likely. With a binary terminal reward, PPO's critic makes this judgment by estimating the current policy's probability of success, but only for states the agent reaches. Our method, WM-PPO, elegantly brings the two together. For each action the policy samples, executed or not, the world model predicts the next observation and the critic values the predicted successor. In the policy channel, comparisons among these values give the policy auxiliary advantages; in the value channel, their mean gives the critic an auxiliary target. A shared gate weights both channels according to how reliably comparisons with similar margins predicted real outcomes earlier in training. The policy, critic and world model all learn from the agent's own interaction, with no teacher or privileged information, and only the policy is used at inference. With a comparable number of real environment steps, WM-PPO raises success from PPO's 63.5% to 87.5% on ALFWorld and from 39.3% to 52.9% on ScienceWorld. It matches the strongest PPO refinements we tested on ALFWorld and exceeds the best of them by 6.5 points on ScienceWorld. The trained policy more often completes prerequisite steps before placing an object or committing to an answer.

Figures & tables

Appendix figures & tables30 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Enhancing Policy Learning with World-Action Model

    Mar 30, 2026Yuci Han, Alper YilmazRobot Policy LearningWorld Action Models

  2. How do World Models and Policies Compose in LLM Agents? A Joint Spectral and Behavioral Account

    Aug 30, 2026Ruize Xu, Xiao Yu, Yujin Tang +2World Model LearningWorld Models

  3. Does Learning to Predict the World Help Agents Act? Auditing World-Model Post-Training

    Sep 27, 2026Xinyu Che, Hang Yan, Yanchen Liu +5World Model LearningRL Post-Training