Robots that follow persistent language instructions face a problem: the world changes, but their cached knowledge about a room may not. The new MIND framework, described in an arXiv preprint, is designed to address this by helping robots determine which cached room judgments are still valid as their environment shifts.

MIND stands for Maintenance via INstruction-conditioned Dependencies. The core idea is to use the instructions themselves as a guide for what to re-check or update, rather than re-evaluating everything. The abstract introduces the framework but does not yet provide experimental results or implementation details.

Because the source is a single short abstract, the description is limited to the problem statement and the proposed approach. No comparisons to other methods or performance claims are available in the provided text.