Search papers, labs, and topics across Lattice.
This paper introduces ActSWM, an action-sensitive latent world model designed to enhance long-horizon planning in open-world games by addressing the issue of Context Collapse, where different action sequences yield indistinguishable future states. By enforcing a transition-separation principle, ActSWM ensures that predicted futures remain responsive to planned actions, thereby improving the stability and accuracy of long-horizon rollouts. Experimental results demonstrate that ActSWM significantly outperforms existing models in maintaining action-dependent differences and achieving higher task success rates in complex interactive environments like Minecraft.
Action-sensitive world models can drastically improve long-horizon planning by preserving critical distinctions between futures generated from different action sequences.
Latent world models support efficient model-predictive control by optimizing future control sequences in latent space and replanning in a receding-horizon manner. However, existing latent predictors often lack stable long-horizon rollout ability, and prediction accuracy alone does not ensure that rollouts remain responsive to the actions being planned. We identify Context Collapse, a failure mode in which autoregressive latent predictors maintain high similarity to future states while producing nearly indistinguishable futures under different action sequences. To address this issue, we propose ActSWM, an action-sensitive latent world model grounded in a transition-separation principle: a planning-useful latent dynamics model should keep alternative-action futures distinguishable and make the action associated with each local transition recoverable. Under this principle, action sensitivity is enforced as a constraint on latent rollouts rather than treated only as an auxiliary prediction target, encouraging predicted futures to preserve action-dependent differences over long horizons. Across step-drift analysis, closed-loop Minecraft planning, and cross-game local action recovery, ActSWM preserves larger action-dependent rollout gaps than existing baselines, improves task success in long-horizon interactive settings, and enables world-model-based action recovery from offline gameplay videos.
