DRAM: Delta-rule Recurrent Associative Memory for Robot Manipulation Policies
Quick summary
arXiv:2609.32453v2 Announce Type: replace-cross Abstract: Robotic manipulation is inherently history-dependent, yet most pretrained robotic policies condition on only the current observation or a short temporal window. Equipping such policies with long-term memory remains challenging: existing approaches either feed the backbone multi-frame observation windows, which substantially increase inference cost, or rely on pre-defined semantic features, which limit task generality and may also require the retraining of the backbone to adapt to the memory. We introduce DRAM (Delta-rule Recurrent Assoc
Key takeaways
- arXiv:2609.32453v2 Announce Type: replace-cross Abstract: Robotic manipulation is inherently history-dependent, yet most pretrained robotic policies condition on only the current observation or a short temporal window.
- Equipping such policies with long-term memory remains challenging: existing approaches either feed the backbone multi-frame observation windows, which substantially increase inference cost, or rely on pre-defined semantic features, which limit task generality and may also require the retraining of the backbone to adapt to the memory.
- We introduce DRAM (Delta-rule Recurrent Assoc
Why it matters
The importance of “DRAM: Delta-rule Recurrent Associative Memory for Robot Manipulation Policies” will be measured by what changes in practice. User behavior, access conditions, verifiable performance and responsible-use outcomes are the signals worth following.
Member comments