cs.AINov 18, 2026

F3^3: What Makes Good Experience in World-Action Models?

Authors: Zihe Huang, Sirui Zhou, Haozhi Han, Ruge Zhang, Teng Ma, Tao Ma, Zheng Liu, Ce Zhang, 

Organizations: Peking University · Institute for AI Industry Research (AIR), Tsinghua University · Independent Researcher · University of Chinese Academy of Sciences · 

Abstract

Deployed world-action models (WAMs) can keep improving only if they make good use of their own experience. Existing methods supply retrieved rollouts or demonstrations to the policy as context, yet such experience often turns previously solved tasks into failures. What, then, makes good experience for a WAM? We show that experience is valuable only in use: its value depends not on whether the source rollout succeeded but on what is stored, when it is used, and where it enters the policy. We therefore propose F3^3, which makes these three decisions with Form, Fit, and Fusion. Form compresses successful and failed rollouts alike into short, verified operation records; Fit teaches the policy to read a record and to predict when it helps; Fusion injects the record's images, text and actions into the policy as a prefix and an action prior. Trained once on only 200 examples and deployed with a growing experience library, the model achieves state-of-the-art results on MolmoSpaces and RoboLab without any training on RoboLab: over native Cosmos 3 Nano, it raises success by 12.0 pp and 10.8 pp, reduces steps to success by 12.2% and 11.3%, and cuts regressions by 57%.

Figures & tables

Explore similar work

CardsList
  1. Seeing Through the Illusion: Test-Time Scaling for World Action Models

    Nov 2, 2026Cheikh Ahmed, Mahdi Mostajabdaveh, Milad Yazdani +1World Models for RoboticsEfficient Inference for World Action Models

  2. Direct Experience World-Model Optimization: Learning the World Beyond Action Imitation

    Sep 29, 2026Xiangcheng Zhan, Zirui Chen, Yicheng Zhao +2Test-Time AdaptationWorld Model Learning

  3. R2R^2-WAM: Repair-and-Reject Post-Training for World Action Models

    Oct 4, 2026Ruiyan Xu, Haisheng Su, Sixu Lin +4World Action ModelsAction-Conditioned World Models