The authors built a fixed-capacity recurrent memory divided into slots.
Spatially addressed memory preserves scenes across long generated videos
SMART uses camera pose to read and write a fixed-capacity memory, helping a video model reproduce previously visited regions without storing an expanding 3D map.
Big Tech
Eren Cetin · Julius Erbach · Vukasin Bozic · Hyojun Go · Zixiang Zhao · Nikolai Kalischek · +3 more
ETH Zurich · Google
Research Digest··3 min read
Cetin et al.
Why this paper
From Google and ETH Zurich
In one line
SMART uses pose-addressed slots for bounded video memory, enabling long-horizon spatial recall without explicit 3D reconstruction.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (2 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§