Event
ASPEN: Predicting Command Persistence to Reduce Redundant VLA Inference
1 reports · 1 sources
Vieri N. Akustia and Bin Hu propose ASPEN, an event-triggered scheduler that reduces repeated semantic inference calls in navigation VLA policies by predicting whether a frozen base policy would reissue the previous command and reusing commands at eligible command boundaries. On frozen NaVILA, evaluated across the RxR-CE val-unseen dataset, ASPEN reduces VLA calls by 9.5%, with an estimated 8.1% reduction in inference time (including scheduling overhead), and maintains action sequences exactly identical to the baseline in 92.1% of episodes, with comparable navigation performance.
Report timeline
- ASPEN: Adaptive Semantic Reasoning via Command Persistence Prediction to Improve Navigation VLA Efficiency
ASPEN is an event-triggered scheduler based on command persistence prediction, designed to reduce repeated semantic reasoning calls in navigation VLA policies. On frozen NaVILA in the cross-dataset RxR-CE val-unseen evaluation, ASPEN reduces VLA calls by 9.5% and estimated inference time by 8.1% (including scheduling overhead), while preserving action sequences identical to the baseline in 92.1% of episodes.
arXiv cs.RO ·