Frozen vision-language-action (VLA) policies cannot incorporate previously encountered failures without updating the underlying model. To address this, the authors propose retrieval-triggered projection correction (RTPC), an auditable external-memory wrapper that retrieves stress-collected local-depth records and applies translation-only projection correction while keeping rotation and gripper commands unchanged.
RTPC was evaluated on CALVIN single-task manipulation using primary and independent replication cohorts. In the primary evaluation, RTPC reduced a depth-based near-collision or stall proxy event rate without changing task-success count. However, this reduction was not reproduced in the independent replication, and retrieved correction directions were not distinguishable from random same-family directions.
The authors identify endpoint coupling, batch sensitivity, and the lack of a trigger-matched memory-free baseline as limitations for attributing effects to retrieval-based correction. These findings do not establish physical safety improvement or memory-specific benefits, but they provide an empirical evaluation and failure analysis of retrieval-conditioned intervention for frozen VLA policies.