AI RESEARCH
DISC: Decoupling Instruction from State-Conditioned Control via Policy Generation
arXiv CS.LG
•
ArXi:2605.20856v1 Announce Type: cross Language-conditioned manipulation policies typically process instructions and observations through shared network parameters. This task-state entanglement provides a pathway for observation leakage -- networks learn scene-to-action shortcuts that bypass language grounding entirely. DISC eliminates this failure structurally. Rather than conditioning a universal policy on language, DISC uses a hypernetwork to generate the entire parameter set of a task-specific visuomotor policy from the instruction alone.