The authors propose ReCAT, a structured recurrent memory policy for language-conditioned manipulation.
Recurrent memory policy improves robot manipulation on tasks requiring spatial recall, counting, and timing
ReCAT combines structured recurrent memory with direct current observation to achieve high success on benchmark and real-world tasks
Big Tech
Pankhuri Vanjani · Mostafa Hatab · Can Mizrakli · Vaisakh Shaj · Zhuoyue Li · Moritz Reuss · +1 more
University of Edinburgh · Karlsruhe Institute of Technology · NVIDIA · Robotics Institute Germany
Research Digest··3 min read
Vanjani et al.
Why this paper
From NVIDIA and 3 others
In one line
ReCAT uses structured recurrent memory with Mamba-2 and causal attention to achieve high success on memory-dependent robot manipulation tasks.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks (4 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§