World Action Models (WAMs) jointly predict robot actions and future observations, but their visual prediction objectives do not explicitly distinguish task-relevant transitions from incidental scene changes. For whole-body humanoid policies, this distinction is critical because task progress depends on coordinated changes in body pose, contact, and object state.
Language-Forcing addresses this by using a frozen vision-language model to assess task progress from paired current and ground-truth future frames, generating descriptions of plausible state changes. These descriptions are encoded and aligned with intermediate action-conditioned future-visual features, complementing the original visual and action prediction objectives. All auxiliary components are used only during training, so there is no additional inference overhead.
On the HumanoidArena benchmark, Language-Forcing achieved an average success rate of 79.17%, outperforming the strongest competing policy by 9.46 percentage points and the same WAM without language alignment by 10.77 points. Across five real-world tasks, language alignment raised average success from 53.33% to 71.33% over the same baseline.