A value-scored hierarchical memory, initialized offline and refined by online rollouts, improves frozen vision-language-action policies on long-horizon manipulation tasks by a few percentage points.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
OnEvoMemory: Evolving Memory through Online Robot Rollouts for Pretrained Robot Policies
A value-scored hierarchical memory, initialized offline and refined by online rollouts, improves frozen vision-language-action policies on long-horizon manipulation tasks by a few percentage points.