FedCausalCompose introduces a causal world-model framework for modular agents to navigate complex service interfaces like payment and inventory. By using local actions as intervention-response evidence, the model overcomes the interventional errors found in standard observational world models.
HOW THIS AFFECTS YOU
●
researcherThis addresses the gap between observational traces and the intervention-time planning required for multi-module systems.