MemoryWAM: Efficient World Action Modeling with Persistent Memory
- Published
- Source
- arXiv
- Paper number
- 456
- Field
- Robotics
- arXiv ID
- 2606.20562
Key points
- The memory design combines a short-term sliding window, long-term gist tokens, and anchor frames at event boundaries.
- It reduces time and space complexity from O(N) to O(N/d), where d is the compression ratio, making real-time control feasible.
- On RMBench, it reaches a 83.0 percent average across nine tasks, outperforming LingBot-VA by 4.8 points and FastWAM by a wide margin.
- On the real robot, it achieves 18 out of 20 on Shell Game and 15 out of 20 on Look and Press.
- Removing gist tokens causes the largest performance drop, from 92.5 percent to 40 percent, which shows their critical role in long-term memory.
- Even at 1,600 frames, it remains more efficient than full attention approaches such as RNN and TTT while maintaining 87 percent performance.
Paper links
External research summaries. These are not HDATF publications or measured product results.